opencv/modules/imgproc/src/smooth.cpp

/*M///////////////////////////////////////////////////////////////////////////////////////
//
//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
//  By downloading, copying, installing or using the software you agree to this license.
//  If you do not agree to this license, do not download, install,
//  copy or use the software.
//
//
//                           License Agreement
//                For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
//   * Redistribution's of source code must retain the above copyright notice,
//     this list of conditions and the following disclaimer.
//
//   * Redistribution's in binary form must reproduce the above copyright notice,
//     this list of conditions and the following disclaimer in the documentation
//     and/or other materials provided with the distribution.
//
//   * The name of the copyright holders may not be used to endorse or promote products
//     derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/

#include "precomp.hpp"
#include "opencl_kernels.hpp"

/*
 * This file includes the code, contributed by Simon Perreault
 * (the function icvMedianBlur_8u_O1)
 *
 * Constant-time median filtering -- http://nomis80.org/ctmf.html
 * Copyright (C) 2006 Simon Perreault
 *
 * Contact:
 *  Laboratoire de vision et systemes numeriques
 *  Pavillon Adrien-Pouliot
 *  Universite Laval
 *  Sainte-Foy, Quebec, Canada
 *  G1K 7P4
 *
 *  perreaul@gel.ulaval.ca
 */

namespace cv
{

/****************************************************************************************\
                                         Box Filter
\****************************************************************************************/

template<typename T, typename ST>
struct RowSum :
        public BaseRowFilter
{
    RowSum( int _ksize, int _anchor ) :
        BaseRowFilter()
    {
        ksize = _ksize;
        anchor = _anchor;
    }

    virtual void operator()(const uchar* src, uchar* dst, int width, int cn)
    {
        const T* S = (const T*)src;
        ST* D = (ST*)dst;
        int i = 0, k, ksz_cn = ksize*cn;

        width = (width - 1)*cn;
        for( k = 0; k < cn; k++, S++, D++ )
        {
            ST s = 0;
            for( i = 0; i < ksz_cn; i += cn )
                s += S[i];
            D[0] = s;
            for( i = 0; i < width; i += cn )
            {
                s += S[i + ksz_cn] - S[i];
                D[i+cn] = s;
            }
        }
    }
};


template<typename ST, typename T>
struct ColumnSum :
        public BaseColumnFilter
{
    ColumnSum( int _ksize, int _anchor, double _scale ) :
        BaseColumnFilter()
    {
        ksize = _ksize;
        anchor = _anchor;
        scale = _scale;
        sumCount = 0;
    }

    virtual void reset() { sumCount = 0; }

    virtual void operator()(const uchar** src, uchar* dst, int dststep, int count, int width)
    {
        int i;
        ST* SUM;
        bool haveScale = scale != 1;
        double _scale = scale;

        if( width != (int)sum.size() )
        {
            sum.resize(width);
            sumCount = 0;
        }

        SUM = &sum[0];
        if( sumCount == 0 )
        {
            for( i = 0; i < width; i++ )
                SUM[i] = 0;
            for( ; sumCount < ksize - 1; sumCount++, src++ )
            {
                const ST* Sp = (const ST*)src[0];
                for( i = 0; i <= width - 2; i += 2 )
                {
                    ST s0 = SUM[i] + Sp[i], s1 = SUM[i+1] + Sp[i+1];
                    SUM[i] = s0; SUM[i+1] = s1;
                }

                for( ; i < width; i++ )
                    SUM[i] += Sp[i];
            }
        }
        else
        {
            CV_Assert( sumCount == ksize-1 );
            src += ksize-1;
        }

        for( ; count--; src++ )
        {
            const ST* Sp = (const ST*)src[0];
            const ST* Sm = (const ST*)src[1-ksize];
            T* D = (T*)dst;
            if( haveScale )
            {
                for( i = 0; i <= width - 2; i += 2 )
                {
                    ST s0 = SUM[i] + Sp[i], s1 = SUM[i+1] + Sp[i+1];
                    D[i] = saturate_cast<T>(s0*_scale);
                    D[i+1] = saturate_cast<T>(s1*_scale);
                    s0 -= Sm[i]; s1 -= Sm[i+1];
                    SUM[i] = s0; SUM[i+1] = s1;
                }

                for( ; i < width; i++ )
                {
                    ST s0 = SUM[i] + Sp[i];
                    D[i] = saturate_cast<T>(s0*_scale);
                    SUM[i] = s0 - Sm[i];
                }
            }
            else
            {
                for( i = 0; i <= width - 2; i += 2 )
                {
                    ST s0 = SUM[i] + Sp[i], s1 = SUM[i+1] + Sp[i+1];
                    D[i] = saturate_cast<T>(s0);
                    D[i+1] = saturate_cast<T>(s1);
                    s0 -= Sm[i]; s1 -= Sm[i+1];
                    SUM[i] = s0; SUM[i+1] = s1;
                }

                for( ; i < width; i++ )
                {
                    ST s0 = SUM[i] + Sp[i];
                    D[i] = saturate_cast<T>(s0);
                    SUM[i] = s0 - Sm[i];
                }
            }
            dst += dststep;
        }
    }

    double scale;
    int sumCount;
    std::vector<ST> sum;
};


template<>
struct ColumnSum<int, uchar> :
        public BaseColumnFilter
{
    ColumnSum( int _ksize, int _anchor, double _scale ) :
        BaseColumnFilter()
    {
        ksize = _ksize;
        anchor = _anchor;
        scale = _scale;
        sumCount = 0;
    }

    virtual void reset() { sumCount = 0; }

    virtual void operator()(const uchar** src, uchar* dst, int dststep, int count, int width)
    {
        int i;
        int* SUM;
        bool haveScale = scale != 1;
        double _scale = scale;

        #if CV_SSE2
            bool haveSSE2 = checkHardwareSupport(CV_CPU_SSE2);
        #endif

        if( width != (int)sum.size() )
        {
            sum.resize(width);
            sumCount = 0;
        }

        SUM = &sum[0];
        if( sumCount == 0 )
        {
            memset((void*)SUM, 0, width*sizeof(int));
            for( ; sumCount < ksize - 1; sumCount++, src++ )
            {
                const int* Sp = (const int*)src[0];
                i = 0;
                #if CV_SSE2
                if(haveSSE2)
                {
                    for( ; i < width-4; i+=4 )
                    {
                        __m128i _sum = _mm_loadu_si128((const __m128i*)(SUM+i));
                        __m128i _sp = _mm_loadu_si128((const __m128i*)(Sp+i));
                        _mm_storeu_si128((__m128i*)(SUM+i),_mm_add_epi32(_sum, _sp));
                    }
                }
                #endif
                for( ; i < width; i++ )
                    SUM[i] += Sp[i];
            }
        }
        else
        {
            CV_Assert( sumCount == ksize-1 );
            src += ksize-1;
        }

        for( ; count--; src++ )
        {
            const int* Sp = (const int*)src[0];
            const int* Sm = (const int*)src[1-ksize];
            uchar* D = (uchar*)dst;
            if( haveScale )
            {
                i = 0;
                #if CV_SSE2
                if(haveSSE2)
                {
                    const __m128 scale4 = _mm_set1_ps((float)_scale);
                    for( ; i < width-8; i+=8 )
                    {
                        __m128i _sm  = _mm_loadu_si128((const __m128i*)(Sm+i));
                        __m128i _sm1  = _mm_loadu_si128((const __m128i*)(Sm+i+4));

                        __m128i _s0  = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i)),
                                                     _mm_loadu_si128((const __m128i*)(Sp+i)));
                        __m128i _s01  = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i+4)),
                                                      _mm_loadu_si128((const __m128i*)(Sp+i+4)));

                        __m128i _s0T = _mm_cvtps_epi32(_mm_mul_ps(scale4, _mm_cvtepi32_ps(_s0)));
                        __m128i _s0T1 = _mm_cvtps_epi32(_mm_mul_ps(scale4, _mm_cvtepi32_ps(_s01)));

                        _s0T = _mm_packs_epi32(_s0T, _s0T1);

                        _mm_storel_epi64((__m128i*)(D+i), _mm_packus_epi16(_s0T, _s0T));

                        _mm_storeu_si128((__m128i*)(SUM+i), _mm_sub_epi32(_s0,_sm));
                        _mm_storeu_si128((__m128i*)(SUM+i+4),_mm_sub_epi32(_s01,_sm1));
                    }
                }
                #endif
                for( ; i < width; i++ )
                {
                    int s0 = SUM[i] + Sp[i];
                    D[i] = saturate_cast<uchar>(s0*_scale);
                    SUM[i] = s0 - Sm[i];
                }
            }
            else
            {
                i = 0;
                #if CV_SSE2
                if(haveSSE2)
                {
                    for( ; i < width-8; i+=8 )
                    {
                        __m128i _sm  = _mm_loadu_si128((const __m128i*)(Sm+i));
                        __m128i _sm1  = _mm_loadu_si128((const __m128i*)(Sm+i+4));

                        __m128i _s0  = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i)),
                                                     _mm_loadu_si128((const __m128i*)(Sp+i)));
                        __m128i _s01  = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i+4)),
                                                      _mm_loadu_si128((const __m128i*)(Sp+i+4)));

                        __m128i _s0T = _mm_packs_epi32(_s0, _s01);

                        _mm_storel_epi64((__m128i*)(D+i), _mm_packus_epi16(_s0T, _s0T));

                        _mm_storeu_si128((__m128i*)(SUM+i), _mm_sub_epi32(_s0,_sm));
                        _mm_storeu_si128((__m128i*)(SUM+i+4),_mm_sub_epi32(_s01,_sm1));
                    }
                }
                #endif

                for( ; i < width; i++ )
                {
                    int s0 = SUM[i] + Sp[i];
                    D[i] = saturate_cast<uchar>(s0);
                    SUM[i] = s0 - Sm[i];
                }
            }
            dst += dststep;
        }
    }

    double scale;
    int sumCount;
    std::vector<int> sum;
};

template<>
struct ColumnSum<int, short> :
        public BaseColumnFilter
{
    ColumnSum( int _ksize, int _anchor, double _scale ) :
        BaseColumnFilter()
    {
        ksize = _ksize;
        anchor = _anchor;
        scale = _scale;
        sumCount = 0;
    }

    virtual void reset() { sumCount = 0; }

    virtual void operator()(const uchar** src, uchar* dst, int dststep, int count, int width)
    {
        int i;
        int* SUM;
        bool haveScale = scale != 1;
        double _scale = scale;

        #if CV_SSE2
            bool haveSSE2 = checkHardwareSupport(CV_CPU_SSE2);
        #endif

        if( width != (int)sum.size() )
        {
            sum.resize(width);
            sumCount = 0;
        }
        SUM = &sum[0];
        if( sumCount == 0 )
        {
            memset((void*)SUM, 0, width*sizeof(int));
            for( ; sumCount < ksize - 1; sumCount++, src++ )
            {
                const int* Sp = (const int*)src[0];
                i = 0;
                #if CV_SSE2
                if(haveSSE2)
                {
                    for( ; i < width-4; i+=4 )
                    {
                        __m128i _sum = _mm_loadu_si128((const __m128i*)(SUM+i));
                        __m128i _sp = _mm_loadu_si128((const __m128i*)(Sp+i));
                        _mm_storeu_si128((__m128i*)(SUM+i),_mm_add_epi32(_sum, _sp));
                    }
                }
                #endif
                for( ; i < width; i++ )
                    SUM[i] += Sp[i];
            }
        }
        else
        {
            CV_Assert( sumCount == ksize-1 );
            src += ksize-1;
        }

        for( ; count--; src++ )
        {
            const int* Sp = (const int*)src[0];
            const int* Sm = (const int*)src[1-ksize];
            short* D = (short*)dst;
            if( haveScale )
            {
                i = 0;
                #if CV_SSE2
                if(haveSSE2)
                {
                    const __m128 scale4 = _mm_set1_ps((float)_scale);
                    for( ; i < width-8; i+=8 )
                    {
                        __m128i _sm   = _mm_loadu_si128((const __m128i*)(Sm+i));
                        __m128i _sm1  = _mm_loadu_si128((const __m128i*)(Sm+i+4));

                        __m128i _s0  = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i)),
                                                     _mm_loadu_si128((const __m128i*)(Sp+i)));
                        __m128i _s01  = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i+4)),
                                                      _mm_loadu_si128((const __m128i*)(Sp+i+4)));

                        __m128i _s0T  = _mm_cvtps_epi32(_mm_mul_ps(scale4, _mm_cvtepi32_ps(_s0)));
                        __m128i _s0T1 = _mm_cvtps_epi32(_mm_mul_ps(scale4, _mm_cvtepi32_ps(_s01)));

                        _mm_storeu_si128((__m128i*)(D+i), _mm_packs_epi32(_s0T, _s0T1));

                        _mm_storeu_si128((__m128i*)(SUM+i),_mm_sub_epi32(_s0,_sm));
                        _mm_storeu_si128((__m128i*)(SUM+i+4), _mm_sub_epi32(_s01,_sm1));
                    }
                }
                #endif
                for( ; i < width; i++ )
                {
                    int s0 = SUM[i] + Sp[i];
                    D[i] = saturate_cast<short>(s0*_scale);
                    SUM[i] = s0 - Sm[i];
                }
            }
            else
            {
                i = 0;
                #if CV_SSE2
                if(haveSSE2)
                {
                    for( ; i < width-8; i+=8 )
                    {

                        __m128i _sm  = _mm_loadu_si128((const __m128i*)(Sm+i));
                        __m128i _sm1  = _mm_loadu_si128((const __m128i*)(Sm+i+4));

                        __m128i _s0  = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i)),
                                                     _mm_loadu_si128((const __m128i*)(Sp+i)));
                        __m128i _s01  = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i+4)),
                                                      _mm_loadu_si128((const __m128i*)(Sp+i+4)));

                        _mm_storeu_si128((__m128i*)(D+i), _mm_packs_epi32(_s0, _s01));

                        _mm_storeu_si128((__m128i*)(SUM+i), _mm_sub_epi32(_s0,_sm));
                        _mm_storeu_si128((__m128i*)(SUM+i+4),_mm_sub_epi32(_s01,_sm1));
                    }
                }
                #endif

                for( ; i < width; i++ )
                {
                    int s0 = SUM[i] + Sp[i];
                    D[i] = saturate_cast<short>(s0);
                    SUM[i] = s0 - Sm[i];
                }
            }
            dst += dststep;
        }
    }

    double scale;
    int sumCount;
    std::vector<int> sum;
};


template<>
struct ColumnSum<int, ushort> :
        public BaseColumnFilter
{
    ColumnSum( int _ksize, int _anchor, double _scale ) :
        BaseColumnFilter()
    {
        ksize = _ksize;
        anchor = _anchor;
        scale = _scale;
        sumCount = 0;
    }

    virtual void reset() { sumCount = 0; }

    virtual void operator()(const uchar** src, uchar* dst, int dststep, int count, int width)
    {
        int i;
        int* SUM;
        bool haveScale = scale != 1;
        double _scale = scale;
        #if CV_SSE2
                bool haveSSE2 =  checkHardwareSupport(CV_CPU_SSE2);
        #endif

        if( width != (int)sum.size() )
        {
            sum.resize(width);
            sumCount = 0;
        }
        SUM = &sum[0];
        if( sumCount == 0 )
        {
            memset((void*)SUM, 0, width*sizeof(int));
            for( ; sumCount < ksize - 1; sumCount++, src++ )
            {
                const int* Sp = (const int*)src[0];
                i = 0;
                #if CV_SSE2
                if(haveSSE2)
                {
                    for( ; i < width-4; i+=4 )
                    {
                        __m128i _sum = _mm_loadu_si128((const __m128i*)(SUM+i));
                        __m128i _sp = _mm_loadu_si128((const __m128i*)(Sp+i));
                        _mm_storeu_si128((__m128i*)(SUM+i), _mm_add_epi32(_sum, _sp));
                    }
                }
                #endif
                for( ; i < width; i++ )
                    SUM[i] += Sp[i];
            }
        }
        else
        {
            CV_Assert( sumCount == ksize-1 );
            src += ksize-1;
        }

        for( ; count--; src++ )
        {
            const int* Sp = (const int*)src[0];
            const int* Sm = (const int*)src[1-ksize];
            ushort* D = (ushort*)dst;
            if( haveScale )
            {
                i = 0;
                #if CV_SSE2
                if(haveSSE2)
                {
                    const __m128 scale4 = _mm_set1_ps((float)_scale);
                    const __m128i delta0 = _mm_set1_epi32(0x8000);
                    const __m128i delta1 = _mm_set1_epi32(0x80008000);

                    for( ; i < width-4; i+=4)
                    {
                        __m128i _sm   = _mm_loadu_si128((const __m128i*)(Sm+i));
                        __m128i _s0   = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i)),
                                                      _mm_loadu_si128((const __m128i*)(Sp+i)));

                        __m128i _res = _mm_cvtps_epi32(_mm_mul_ps(scale4, _mm_cvtepi32_ps(_s0)));

                        _res = _mm_sub_epi32(_res, delta0);
                        _res = _mm_add_epi16(_mm_packs_epi32(_res, _res), delta1);

                        _mm_storel_epi64((__m128i*)(D+i), _res);
                        _mm_storeu_si128((__m128i*)(SUM+i), _mm_sub_epi32(_s0,_sm));
                    }
                }
                #endif
                for( ; i < width; i++ )
                {
                    int s0 = SUM[i] + Sp[i];
                    D[i] = saturate_cast<ushort>(s0*_scale);
                    SUM[i] = s0 - Sm[i];
                }
            }
            else
            {
                i = 0;
                #if  CV_SSE2
                if(haveSSE2)
                {
                    const __m128i delta0 = _mm_set1_epi32(0x8000);
                    const __m128i delta1 = _mm_set1_epi32(0x80008000);

                    for( ; i < width-4; i+=4 )
                    {
                        __m128i _sm   = _mm_loadu_si128((const __m128i*)(Sm+i));
                        __m128i _s0   = _mm_add_epi32(_mm_loadu_si128((const __m128i*)(SUM+i)),
                                                      _mm_loadu_si128((const __m128i*)(Sp+i)));

                        __m128i _res = _mm_sub_epi32(_s0, delta0);
                        _res = _mm_add_epi16(_mm_packs_epi32(_res, _res), delta1);

                        _mm_storel_epi64((__m128i*)(D+i), _res);
                        _mm_storeu_si128((__m128i*)(SUM+i), _mm_sub_epi32(_s0,_sm));
                    }
                }
                #endif

                for( ; i < width; i++ )
                {
                    int s0 = SUM[i] + Sp[i];
                    D[i] = saturate_cast<ushort>(s0);
                    SUM[i] = s0 - Sm[i];
                }
            }
            dst += dststep;
        }
    }

    double scale;
    int sumCount;
    std::vector<int> sum;
};

#ifdef HAVE_OPENCL

#define DIVUP(total, grain) ((total + grain - 1) / (grain))

static bool ocl_boxFilter( InputArray _src, OutputArray _dst, int ddepth,
                           Size ksize, Point anchor, int borderType, bool normalize, bool sqr = false )
{
    int type = _src.type(), sdepth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type), esz = CV_ELEM_SIZE(type);
    bool doubleSupport = ocl::Device::getDefault().doubleFPConfig() > 0;

    if (ddepth < 0)
        ddepth = sdepth;

    if (cn > 4 || (!doubleSupport && (sdepth == CV_64F || ddepth == CV_64F)) ||
        _src.offset() % esz != 0 || _src.step() % esz != 0)
        return false;

    if (anchor.x < 0)
        anchor.x = ksize.width / 2;
    if (anchor.y < 0)
        anchor.y = ksize.height / 2;

    int computeUnits = ocl::Device::getDefault().maxComputeUnits();
    float alpha = 1.0f / (ksize.height * ksize.width);
    Size size = _src.size(), wholeSize;
    bool isolated = (borderType & BORDER_ISOLATED) != 0;
    borderType &= ~BORDER_ISOLATED;
    int wdepth = std::max(CV_32F, std::max(ddepth, sdepth));

    const char * const borderMap[] = { "BORDER_CONSTANT", "BORDER_REPLICATE", "BORDER_REFLECT", 0, "BORDER_REFLECT_101" };
    size_t globalsize[2] = { size.width, size.height };
    size_t localsize[2] = { 0, 1 };

    UMat src = _src.getUMat();
    if (!isolated)
    {
        Point ofs;
        src.locateROI(wholeSize, ofs);
    }

    int h = isolated ? size.height : wholeSize.height;
    int w = isolated ? size.width : wholeSize.width;

    size_t maxWorkItemSizes[32];
    ocl::Device::getDefault().maxWorkItemSizes(maxWorkItemSizes);
    int tryWorkItems = (int)maxWorkItemSizes[0];

    ocl::Kernel kernel;
    for ( ; ; )
    {
        int BLOCK_SIZE_X = tryWorkItems, BLOCK_SIZE_Y = std::min(ksize.height * 10, size.height);

        while (BLOCK_SIZE_X > 32 && BLOCK_SIZE_X >= ksize.width * 2 && BLOCK_SIZE_X > size.width * 2)
            BLOCK_SIZE_X /= 2;
        while (BLOCK_SIZE_Y < BLOCK_SIZE_X / 8 && BLOCK_SIZE_Y * computeUnits * 32 < size.height)
            BLOCK_SIZE_Y *= 2;

        if (ksize.width > BLOCK_SIZE_X || w < ksize.width || h < ksize.height)
            return false;

        char cvt[2][50];
        String opts = format("-D LOCAL_SIZE_X=%d -D BLOCK_SIZE_Y=%d -D ST=%s -D DT=%s -D WT=%s -D convertToDT=%s -D convertToWT=%s"
                             " -D ANCHOR_X=%d -D ANCHOR_Y=%d -D KERNEL_SIZE_X=%d -D KERNEL_SIZE_Y=%d -D %s%s%s%s%s"
                             " -D ST1=%s -D DT1=%s -D cn=%d",
                             BLOCK_SIZE_X, BLOCK_SIZE_Y, ocl::typeToStr(type), ocl::typeToStr(CV_MAKE_TYPE(ddepth, cn)),
                             ocl::typeToStr(CV_MAKE_TYPE(wdepth, cn)),
                             ocl::convertTypeStr(wdepth, ddepth, cn, cvt[0]),
                             ocl::convertTypeStr(sdepth, wdepth, cn, cvt[1]),
                             anchor.x, anchor.y, ksize.width, ksize.height, borderMap[borderType],
                             isolated ? " -D BORDER_ISOLATED" : "", doubleSupport ? " -D DOUBLE_SUPPORT" : "",
                             normalize ? " -D NORMALIZE" : "", sqr ? " -D SQR" : "",
                             ocl::typeToStr(sdepth), ocl::typeToStr(ddepth), cn);

        localsize[0] = BLOCK_SIZE_X;
        globalsize[0] = DIVUP(size.width, BLOCK_SIZE_X - (ksize.width - 1)) * BLOCK_SIZE_X;
        globalsize[1] = DIVUP(size.height, BLOCK_SIZE_Y);

        kernel.create("boxFilter", cv::ocl::imgproc::boxFilter_oclsrc, opts);

        size_t kernelWorkGroupSize = kernel.workGroupSize();
        if (localsize[0] <= kernelWorkGroupSize)
            break;
        if (BLOCK_SIZE_X < (int)kernelWorkGroupSize)
            return false;

        tryWorkItems = (int)kernelWorkGroupSize;
    }

    _dst.create(size, CV_MAKETYPE(ddepth, cn));
    UMat dst = _dst.getUMat();

    int idxArg = kernel.set(0, ocl::KernelArg::PtrReadOnly(src));
    idxArg = kernel.set(idxArg, (int)src.step);
    int srcOffsetX = (int)((src.offset % src.step) / src.elemSize());
    int srcOffsetY = (int)(src.offset / src.step);
    int srcEndX = isolated ? srcOffsetX + size.width : wholeSize.width;
    int srcEndY = isolated ? srcOffsetY + size.height : wholeSize.height;
    idxArg = kernel.set(idxArg, srcOffsetX);
    idxArg = kernel.set(idxArg, srcOffsetY);
    idxArg = kernel.set(idxArg, srcEndX);
    idxArg = kernel.set(idxArg, srcEndY);
    idxArg = kernel.set(idxArg, ocl::KernelArg::WriteOnly(dst));
    if (normalize)
        idxArg = kernel.set(idxArg, (float)alpha);

    return kernel.run(2, globalsize, localsize, false);
}

#endif

}


cv::Ptr<cv::BaseRowFilter> cv::getRowSumFilter(int srcType, int sumType, int ksize, int anchor)
{
    int sdepth = CV_MAT_DEPTH(srcType), ddepth = CV_MAT_DEPTH(sumType);
    CV_Assert( CV_MAT_CN(sumType) == CV_MAT_CN(srcType) );

    if( anchor < 0 )
        anchor = ksize/2;

    if( sdepth == CV_8U && ddepth == CV_32S )
        return makePtr<RowSum<uchar, int> >(ksize, anchor);
    if( sdepth == CV_8U && ddepth == CV_64F )
        return makePtr<RowSum<uchar, double> >(ksize, anchor);
    if( sdepth == CV_16U && ddepth == CV_32S )
        return makePtr<RowSum<ushort, int> >(ksize, anchor);
    if( sdepth == CV_16U && ddepth == CV_64F )
        return makePtr<RowSum<ushort, double> >(ksize, anchor);
    if( sdepth == CV_16S && ddepth == CV_32S )
        return makePtr<RowSum<short, int> >(ksize, anchor);
    if( sdepth == CV_32S && ddepth == CV_32S )
        return makePtr<RowSum<int, int> >(ksize, anchor);
    if( sdepth == CV_16S && ddepth == CV_64F )
        return makePtr<RowSum<short, double> >(ksize, anchor);
    if( sdepth == CV_32F && ddepth == CV_64F )
        return makePtr<RowSum<float, double> >(ksize, anchor);
    if( sdepth == CV_64F && ddepth == CV_64F )
        return makePtr<RowSum<double, double> >(ksize, anchor);

    CV_Error_( CV_StsNotImplemented,
        ("Unsupported combination of source format (=%d), and buffer format (=%d)",
        srcType, sumType));

    return Ptr<BaseRowFilter>();
}


cv::Ptr<cv::BaseColumnFilter> cv::getColumnSumFilter(int sumType, int dstType, int ksize,
                                                     int anchor, double scale)
{
    int sdepth = CV_MAT_DEPTH(sumType), ddepth = CV_MAT_DEPTH(dstType);
    CV_Assert( CV_MAT_CN(sumType) == CV_MAT_CN(dstType) );

    if( anchor < 0 )
        anchor = ksize/2;

    if( ddepth == CV_8U && sdepth == CV_32S )
        return makePtr<ColumnSum<int, uchar> >(ksize, anchor, scale);
    if( ddepth == CV_8U && sdepth == CV_64F )
        return makePtr<ColumnSum<double, uchar> >(ksize, anchor, scale);
    if( ddepth == CV_16U && sdepth == CV_32S )
        return makePtr<ColumnSum<int, ushort> >(ksize, anchor, scale);
    if( ddepth == CV_16U && sdepth == CV_64F )
        return makePtr<ColumnSum<double, ushort> >(ksize, anchor, scale);
    if( ddepth == CV_16S && sdepth == CV_32S )
        return makePtr<ColumnSum<int, short> >(ksize, anchor, scale);
    if( ddepth == CV_16S && sdepth == CV_64F )
        return makePtr<ColumnSum<double, short> >(ksize, anchor, scale);
    if( ddepth == CV_32S && sdepth == CV_32S )
        return makePtr<ColumnSum<int, int> >(ksize, anchor, scale);
    if( ddepth == CV_32F && sdepth == CV_32S )
        return makePtr<ColumnSum<int, float> >(ksize, anchor, scale);
    if( ddepth == CV_32F && sdepth == CV_64F )
        return makePtr<ColumnSum<double, float> >(ksize, anchor, scale);
    if( ddepth == CV_64F && sdepth == CV_32S )
        return makePtr<ColumnSum<int, double> >(ksize, anchor, scale);
    if( ddepth == CV_64F && sdepth == CV_64F )
        return makePtr<ColumnSum<double, double> >(ksize, anchor, scale);

    CV_Error_( CV_StsNotImplemented,
        ("Unsupported combination of sum format (=%d), and destination format (=%d)",
        sumType, dstType));

    return Ptr<BaseColumnFilter>();
}


cv::Ptr<cv::FilterEngine> cv::createBoxFilter( int srcType, int dstType, Size ksize,
                    Point anchor, bool normalize, int borderType )
{
    int sdepth = CV_MAT_DEPTH(srcType);
    int cn = CV_MAT_CN(srcType), sumType = CV_64F;
    if( sdepth <= CV_32S && (!normalize ||
        ksize.width*ksize.height <= (sdepth == CV_8U ? (1<<23) :
            sdepth == CV_16U ? (1 << 15) : (1 << 16))) )
        sumType = CV_32S;
    sumType = CV_MAKETYPE( sumType, cn );

    Ptr<BaseRowFilter> rowFilter = getRowSumFilter(srcType, sumType, ksize.width, anchor.x );
    Ptr<BaseColumnFilter> columnFilter = getColumnSumFilter(sumType,
        dstType, ksize.height, anchor.y, normalize ? 1./(ksize.width*ksize.height) : 1);

    return makePtr<FilterEngine>(Ptr<BaseFilter>(), rowFilter, columnFilter,
           srcType, dstType, sumType, borderType );
}


void cv::boxFilter( InputArray _src, OutputArray _dst, int ddepth,
                Size ksize, Point anchor,
                bool normalize, int borderType )
{
    CV_OCL_RUN(_dst.isUMat(), ocl_boxFilter(_src, _dst, ddepth, ksize, anchor, borderType, normalize))

    Mat src = _src.getMat();
    int stype = src.type(), sdepth = CV_MAT_DEPTH(stype), cn = CV_MAT_CN(stype);
    if( ddepth < 0 )
        ddepth = sdepth;
    _dst.create( src.size(), CV_MAKETYPE(ddepth, cn) );
    Mat dst = _dst.getMat();
    if( borderType != BORDER_CONSTANT && normalize && (borderType & BORDER_ISOLATED) != 0 )
    {
        if( src.rows == 1 )
            ksize.height = 1;
        if( src.cols == 1 )
            ksize.width = 1;
    }
#ifdef HAVE_TEGRA_OPTIMIZATION
    if ( tegra::box(src, dst, ksize, anchor, normalize, borderType) )
        return;
#endif

#if defined(HAVE_IPP)
    int ippBorderType = borderType & ~BORDER_ISOLATED;
    Point ocvAnchor, ippAnchor;
    ocvAnchor.x = anchor.x < 0 ? ksize.width / 2 : anchor.x;
    ocvAnchor.y = anchor.y < 0 ? ksize.height / 2 : anchor.y;
    ippAnchor.x = ksize.width / 2 - (ksize.width % 2 == 0 ? 1 : 0);
    ippAnchor.y = ksize.height / 2 - (ksize.height % 2 == 0 ? 1 : 0);

    if (normalize && !src.isSubmatrix() && ddepth == sdepth &&
        (/*ippBorderType == BORDER_REPLICATE ||*/ /* returns ippStsStepErr: Step value is not valid */
         ippBorderType == BORDER_CONSTANT) && ocvAnchor == ippAnchor )
    {
        Ipp32s bufSize = 0;
        IppiSize roiSize = { dst.cols, dst.rows }, maskSize = { ksize.width, ksize.height };

#define IPP_FILTER_BOX_BORDER(ippType, ippDataType, flavor) \
        do \
        { \
            if (ippiFilterBoxBorderGetBufferSize(roiSize, maskSize, ippDataType, cn, &bufSize) >= 0) \
            { \
                Ipp8u * buffer = ippsMalloc_8u(bufSize); \
                ippType borderValue[4] = { 0, 0, 0, 0 }; \
                ippBorderType = ippBorderType == BORDER_CONSTANT ? ippBorderConst : ippBorderRepl; \
                IppStatus status = ippiFilterBoxBorder_##flavor((const ippType *)src.data, (int)src.step, (ippType *)dst.data, \
                                                                (int)dst.step, roiSize, maskSize, \
                                                                (IppiBorderType)ippBorderType, borderValue, buffer); \
                ippsFree(buffer); \
                if (status >= 0) \
                    return; \
            } \
            setIppErrorStatus(); \
        } while ((void)0, 0)

        if (stype == CV_8UC1)
            IPP_FILTER_BOX_BORDER(Ipp8u, ipp8u, 8u_C1R);
        else if (stype == CV_8UC3)
            IPP_FILTER_BOX_BORDER(Ipp8u, ipp8u, 8u_C3R);
        else if (stype == CV_8UC4)
            IPP_FILTER_BOX_BORDER(Ipp8u, ipp8u, 8u_C4R);

        else if (stype == CV_16UC1)
            IPP_FILTER_BOX_BORDER(Ipp16u, ipp16u, 16u_C1R);
        else if (stype == CV_16UC3)
            IPP_FILTER_BOX_BORDER(Ipp16u, ipp16u, 16u_C3R);
        else if (stype == CV_16UC4)
            IPP_FILTER_BOX_BORDER(Ipp16u, ipp16u, 16u_C4R);

        else if (stype == CV_16SC1)
            IPP_FILTER_BOX_BORDER(Ipp16s, ipp16s, 16s_C1R);
        else if (stype == CV_16SC3)
            IPP_FILTER_BOX_BORDER(Ipp16s, ipp16s, 16s_C3R);
        else if (stype == CV_16SC4)
            IPP_FILTER_BOX_BORDER(Ipp16s, ipp16s, 16s_C4R);

        else if (stype == CV_32FC1)
            IPP_FILTER_BOX_BORDER(Ipp32f, ipp32f, 32f_C1R);
        else if (stype == CV_32FC3)
            IPP_FILTER_BOX_BORDER(Ipp32f, ipp32f, 32f_C3R);
        else if (stype == CV_32FC4)
            IPP_FILTER_BOX_BORDER(Ipp32f, ipp32f, 32f_C4R);
    }
#undef IPP_FILTER_BOX_BORDER
#endif

    Ptr<FilterEngine> f = createBoxFilter( src.type(), dst.type(),
                        ksize, anchor, normalize, borderType );
    f->apply( src, dst );
}

void cv::blur( InputArray src, OutputArray dst,
           Size ksize, Point anchor, int borderType )
{
    boxFilter( src, dst, -1, ksize, anchor, true, borderType );
}


/****************************************************************************************\
                                    Squared Box Filter
\****************************************************************************************/

namespace cv
{

template<typename T, typename ST>
struct SqrRowSum :
        public BaseRowFilter
{
    SqrRowSum( int _ksize, int _anchor ) :
        BaseRowFilter()
    {
        ksize = _ksize;
        anchor = _anchor;
    }

    virtual void operator()(const uchar* src, uchar* dst, int width, int cn)
    {
        const T* S = (const T*)src;
        ST* D = (ST*)dst;
        int i = 0, k, ksz_cn = ksize*cn;

        width = (width - 1)*cn;
        for( k = 0; k < cn; k++, S++, D++ )
        {
            ST s = 0;
            for( i = 0; i < ksz_cn; i += cn )
            {
                ST val = (ST)S[i];
                s += val*val;
            }
            D[0] = s;
            for( i = 0; i < width; i += cn )
            {
                ST val0 = (ST)S[i], val1 = (ST)S[i + ksz_cn];
                s += val1*val1 - val0*val0;
                D[i+cn] = s;
            }
        }
    }
};

static Ptr<BaseRowFilter> getSqrRowSumFilter(int srcType, int sumType, int ksize, int anchor)
{
    int sdepth = CV_MAT_DEPTH(srcType), ddepth = CV_MAT_DEPTH(sumType);
    CV_Assert( CV_MAT_CN(sumType) == CV_MAT_CN(srcType) );

    if( anchor < 0 )
        anchor = ksize/2;

    if( sdepth == CV_8U && ddepth == CV_32S )
        return makePtr<SqrRowSum<uchar, int> >(ksize, anchor);
    if( sdepth == CV_8U && ddepth == CV_64F )
        return makePtr<SqrRowSum<uchar, double> >(ksize, anchor);
    if( sdepth == CV_16U && ddepth == CV_64F )
        return makePtr<SqrRowSum<ushort, double> >(ksize, anchor);
    if( sdepth == CV_16S && ddepth == CV_64F )
        return makePtr<SqrRowSum<short, double> >(ksize, anchor);
    if( sdepth == CV_32F && ddepth == CV_64F )
        return makePtr<SqrRowSum<float, double> >(ksize, anchor);
    if( sdepth == CV_64F && ddepth == CV_64F )
        return makePtr<SqrRowSum<double, double> >(ksize, anchor);

    CV_Error_( CV_StsNotImplemented,
              ("Unsupported combination of source format (=%d), and buffer format (=%d)",
               srcType, sumType));

    return Ptr<BaseRowFilter>();
}

}

void cv::sqrBoxFilter( InputArray _src, OutputArray _dst, int ddepth,
                       Size ksize, Point anchor,
                       bool normalize, int borderType )
{
    int srcType = _src.type(), sdepth = CV_MAT_DEPTH(srcType), cn = CV_MAT_CN(srcType);
    Size size = _src.size();

    if( ddepth < 0 )
        ddepth = sdepth < CV_32F ? CV_32F : CV_64F;

    if( borderType != BORDER_CONSTANT && normalize )
    {
        if( size.height == 1 )
            ksize.height = 1;
        if( size.width == 1 )
            ksize.width = 1;
    }

    CV_OCL_RUN(_dst.isUMat() && _src.dims() <= 2,
               ocl_boxFilter(_src, _dst, ddepth, ksize, anchor, borderType, normalize, true))

    int sumDepth = CV_64F;
    if( sdepth == CV_8U )
        sumDepth = CV_32S;
    int sumType = CV_MAKETYPE( sumDepth, cn ), dstType = CV_MAKETYPE(ddepth, cn);

    Mat src = _src.getMat();
    _dst.create( size, dstType );
    Mat dst = _dst.getMat();

    Ptr<BaseRowFilter> rowFilter = getSqrRowSumFilter(srcType, sumType, ksize.width, anchor.x );
    Ptr<BaseColumnFilter> columnFilter = getColumnSumFilter(sumType,
                                                            dstType, ksize.height, anchor.y,
                                                            normalize ? 1./(ksize.width*ksize.height) : 1);

    Ptr<FilterEngine> f = makePtr<FilterEngine>(Ptr<BaseFilter>(), rowFilter, columnFilter,
                                                srcType, dstType, sumType, borderType );
    f->apply( src, dst );
}


/****************************************************************************************\
                                     Gaussian Blur
\****************************************************************************************/

cv::Mat cv::getGaussianKernel( int n, double sigma, int ktype )
{
    const int SMALL_GAUSSIAN_SIZE = 7;
    static const float small_gaussian_tab[][SMALL_GAUSSIAN_SIZE] =
    {
        {1.f},
        {0.25f, 0.5f, 0.25f},
        {0.0625f, 0.25f, 0.375f, 0.25f, 0.0625f},
        {0.03125f, 0.109375f, 0.21875f, 0.28125f, 0.21875f, 0.109375f, 0.03125f}
    };

    const float* fixed_kernel = n % 2 == 1 && n <= SMALL_GAUSSIAN_SIZE && sigma <= 0 ?
        small_gaussian_tab[n>>1] : 0;

    CV_Assert( ktype == CV_32F || ktype == CV_64F );
    Mat kernel(n, 1, ktype);
    float* cf = (float*)kernel.data;
    double* cd = (double*)kernel.data;

    double sigmaX = sigma > 0 ? sigma : ((n-1)*0.5 - 1)*0.3 + 0.8;
    double scale2X = -0.5/(sigmaX*sigmaX);
    double sum = 0;

    int i;
    for( i = 0; i < n; i++ )
    {
        double x = i - (n-1)*0.5;
        double t = fixed_kernel ? (double)fixed_kernel[i] : std::exp(scale2X*x*x);
        if( ktype == CV_32F )
        {
            cf[i] = (float)t;
            sum += cf[i];
        }
        else
        {
            cd[i] = t;
            sum += cd[i];
        }
    }

    sum = 1./sum;
    for( i = 0; i < n; i++ )
    {
        if( ktype == CV_32F )
            cf[i] = (float)(cf[i]*sum);
        else
            cd[i] *= sum;
    }

    return kernel;
}

namespace cv {

static void createGaussianKernels( Mat & kx, Mat & ky, int type, Size ksize,
                                   double sigma1, double sigma2 )
{
    int depth = CV_MAT_DEPTH(type);
    if( sigma2 <= 0 )
        sigma2 = sigma1;

    // automatic detection of kernel size from sigma
    if( ksize.width <= 0 && sigma1 > 0 )
        ksize.width = cvRound(sigma1*(depth == CV_8U ? 3 : 4)*2 + 1)|1;
    if( ksize.height <= 0 && sigma2 > 0 )
        ksize.height = cvRound(sigma2*(depth == CV_8U ? 3 : 4)*2 + 1)|1;

    CV_Assert( ksize.width > 0 && ksize.width % 2 == 1 &&
        ksize.height > 0 && ksize.height % 2 == 1 );

    sigma1 = std::max( sigma1, 0. );
    sigma2 = std::max( sigma2, 0. );

    kx = getGaussianKernel( ksize.width, sigma1, std::max(depth, CV_32F) );
    if( ksize.height == ksize.width && std::abs(sigma1 - sigma2) < DBL_EPSILON )
        ky = kx;
    else
        ky = getGaussianKernel( ksize.height, sigma2, std::max(depth, CV_32F) );
}

}

cv::Ptr<cv::FilterEngine> cv::createGaussianFilter( int type, Size ksize,
                                        double sigma1, double sigma2,
                                        int borderType )
{
    Mat kx, ky;
    createGaussianKernels(kx, ky, type, ksize, sigma1, sigma2);

    return createSeparableLinearFilter( type, type, kx, ky, Point(-1,-1), 0, borderType );
}


void cv::GaussianBlur( InputArray _src, OutputArray _dst, Size ksize,
                   double sigma1, double sigma2,
                   int borderType )
{
    int type = _src.type();
    Size size = _src.size();
    _dst.create( size, type );

    if( borderType != BORDER_CONSTANT && (borderType & BORDER_ISOLATED) != 0 )
    {
        if( size.height == 1 )
            ksize.height = 1;
        if( size.width == 1 )
            ksize.width = 1;
    }

    if( ksize.width == 1 && ksize.height == 1 )
    {
        _src.copyTo(_dst);
        return;
    }

#ifdef HAVE_TEGRA_OPTIMIZATION
    if(sigma1 == 0 && sigma2 == 0 && tegra::gaussian(_src.getMat(), _dst.getMat(), ksize, borderType))
        return;
#endif

#if IPP_VERSION_X100 >= 801 && 0 // these functions are slower in IPP 8.1
    int depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);

    if ((depth == CV_8U || depth == CV_16U || depth == CV_16S || depth == CV_32F) && (cn == 1 || cn == 3) &&
            sigma1 == sigma2 && ksize.width == ksize.height && sigma1 != 0.0 )
    {
        IppiBorderType ippBorder = ippiGetBorderType(borderType);
        if (ippBorderConst == ippBorder || ippBorderRepl == ippBorder)
        {
            Mat src = _src.getMat(), dst = _dst.getMat();
            IppiSize roiSize = { src.cols, src.rows };
            IppDataType dataType = ippiGetDataType(depth);
            Ipp32s specSize = 0, bufferSize = 0;

            if (ippiFilterGaussianGetBufferSize(roiSize, (Ipp32u)ksize.width, dataType, cn, &specSize, &bufferSize) >= 0)
            {
                IppFilterGaussianSpec * pSpec = (IppFilterGaussianSpec *)ippMalloc(specSize);
                Ipp8u * pBuffer = (Ipp8u*)ippMalloc(bufferSize);

                if (ippiFilterGaussianInit(roiSize, (Ipp32u)ksize.width, (Ipp32f)sigma1, ippBorder, dataType, 1, pSpec, pBuffer) >= 0)
                {
#define IPP_FILTER_GAUSS(ippfavor, ippcn) \
    do \
    { \
        typedef Ipp##ippfavor ippType; \
        ippType borderValues[] = { 0, 0, 0 }; \
        IppStatus status = ippcn == 1 ? \
            ippiFilterGaussianBorder_##ippfavor##_C1R((const ippType *)src.data, (int)src.step, \
                (ippType *)dst.data, (int)dst.step, roiSize, borderValues[0], pSpec, pBuffer) : \
            ippiFilterGaussianBorder_##ippfavor##_C3R((const ippType *)src.data, (int)src.step, \
                (ippType *)dst.data, (int)dst.step, roiSize, borderValues, pSpec, pBuffer); \
        ippFree(pBuffer); \
        ippFree(pSpec); \
        if (status >= 0) \
            return; \
    } while ((void)0, 0)

                    if (type == CV_8UC1)
                        IPP_FILTER_GAUSS(8u, 1);
                    else if (type == CV_8UC3)
                        IPP_FILTER_GAUSS(8u, 3);
                    else if (type == CV_16UC1)
                        IPP_FILTER_GAUSS(16u, 1);
                    else if (type == CV_16UC3)
                        IPP_FILTER_GAUSS(16u, 3);
                    else if (type == CV_16SC1)
                        IPP_FILTER_GAUSS(16s, 1);
                    else if (type == CV_16SC3)
                        IPP_FILTER_GAUSS(16s, 3);
                    else if (type == CV_32FC1)
                        IPP_FILTER_GAUSS(32f, 1);
                    else if (type == CV_32FC3)
                        IPP_FILTER_GAUSS(32f, 3);
#undef IPP_FILTER_GAUSS
                }
            }
            setIppErrorStatus();
        }
    }
#endif

    Mat kx, ky;
    createGaussianKernels(kx, ky, type, ksize, sigma1, sigma2);
    sepFilter2D(_src, _dst, CV_MAT_DEPTH(type), kx, ky, Point(-1,-1), 0, borderType );
}

/****************************************************************************************\
                                      Median Filter
\****************************************************************************************/

namespace cv
{
typedef ushort HT;

/**
 * This structure represents a two-tier histogram. The first tier (known as the
 * "coarse" level) is 4 bit wide and the second tier (known as the "fine" level)
 * is 8 bit wide. Pixels inserted in the fine level also get inserted into the
 * coarse bucket designated by the 4 MSBs of the fine bucket value.
 *
 * The structure is aligned on 16 bits, which is a prerequisite for SIMD
 * instructions. Each bucket is 16 bit wide, which means that extra care must be
 * taken to prevent overflow.
 */
typedef struct
{
    HT coarse[16];
    HT fine[16][16];
} Histogram;


#if CV_SSE2
#define MEDIAN_HAVE_SIMD 1

static inline void histogram_add_simd( const HT x[16], HT y[16] )
{
    const __m128i* rx = (const __m128i*)x;
    __m128i* ry = (__m128i*)y;
    __m128i r0 = _mm_add_epi16(_mm_load_si128(ry+0),_mm_load_si128(rx+0));
    __m128i r1 = _mm_add_epi16(_mm_load_si128(ry+1),_mm_load_si128(rx+1));
    _mm_store_si128(ry+0, r0);
    _mm_store_si128(ry+1, r1);
}

static inline void histogram_sub_simd( const HT x[16], HT y[16] )
{
    const __m128i* rx = (const __m128i*)x;
    __m128i* ry = (__m128i*)y;
    __m128i r0 = _mm_sub_epi16(_mm_load_si128(ry+0),_mm_load_si128(rx+0));
    __m128i r1 = _mm_sub_epi16(_mm_load_si128(ry+1),_mm_load_si128(rx+1));
    _mm_store_si128(ry+0, r0);
    _mm_store_si128(ry+1, r1);
}

#else
#define MEDIAN_HAVE_SIMD 0
#endif


static inline void histogram_add( const HT x[16], HT y[16] )
{
    int i;
    for( i = 0; i < 16; ++i )
        y[i] = (HT)(y[i] + x[i]);
}

static inline void histogram_sub( const HT x[16], HT y[16] )
{
    int i;
    for( i = 0; i < 16; ++i )
        y[i] = (HT)(y[i] - x[i]);
}

static inline void histogram_muladd( int a, const HT x[16],
        HT y[16] )
{
    for( int i = 0; i < 16; ++i )
        y[i] = (HT)(y[i] + a * x[i]);
}

static void
medianBlur_8u_O1( const Mat& _src, Mat& _dst, int ksize )
{
/**
 * HOP is short for Histogram OPeration. This macro makes an operation \a op on
 * histogram \a h for pixel value \a x. It takes care of handling both levels.
 */
#define HOP(h,x,op) \
    h.coarse[x>>4] op, \
    *((HT*)h.fine + x) op

#define COP(c,j,x,op) \
    h_coarse[ 16*(n*c+j) + (x>>4) ] op, \
    h_fine[ 16 * (n*(16*c+(x>>4)) + j) + (x & 0xF) ] op

    int cn = _dst.channels(), m = _dst.rows, r = (ksize-1)/2;
    size_t sstep = _src.step, dstep = _dst.step;
    Histogram CV_DECL_ALIGNED(16) H[4];
    HT CV_DECL_ALIGNED(16) luc[4][16];

    int STRIPE_SIZE = std::min( _dst.cols, 512/cn );

    std::vector<HT> _h_coarse(1 * 16 * (STRIPE_SIZE + 2*r) * cn + 16);
    std::vector<HT> _h_fine(16 * 16 * (STRIPE_SIZE + 2*r) * cn + 16);
    HT* h_coarse = alignPtr(&_h_coarse[0], 16);
    HT* h_fine = alignPtr(&_h_fine[0], 16);
#if MEDIAN_HAVE_SIMD
    volatile bool useSIMD = checkHardwareSupport(CV_CPU_SSE2);
#endif

    for( int x = 0; x < _dst.cols; x += STRIPE_SIZE )
    {
        int i, j, k, c, n = std::min(_dst.cols - x, STRIPE_SIZE) + r*2;
        const uchar* src = _src.data + x*cn;
        uchar* dst = _dst.data + (x - r)*cn;

        memset( h_coarse, 0, 16*n*cn*sizeof(h_coarse[0]) );
        memset( h_fine, 0, 16*16*n*cn*sizeof(h_fine[0]) );

        // First row initialization
        for( c = 0; c < cn; c++ )
        {
            for( j = 0; j < n; j++ )
                COP( c, j, src[cn*j+c], += (cv::HT)(r+2) );

            for( i = 1; i < r; i++ )
            {
                const uchar* p = src + sstep*std::min(i, m-1);
                for ( j = 0; j < n; j++ )
                    COP( c, j, p[cn*j+c], ++ );
            }
        }

        for( i = 0; i < m; i++ )
        {
            const uchar* p0 = src + sstep * std::max( 0, i-r-1 );
            const uchar* p1 = src + sstep * std::min( m-1, i+r );

            memset( H, 0, cn*sizeof(H[0]) );
            memset( luc, 0, cn*sizeof(luc[0]) );
            for( c = 0; c < cn; c++ )
            {
                // Update column histograms for the entire row.
                for( j = 0; j < n; j++ )
                {
                    COP( c, j, p0[j*cn + c], -- );
                    COP( c, j, p1[j*cn + c], ++ );
                }

                // First column initialization
                for( k = 0; k < 16; ++k )
                    histogram_muladd( 2*r+1, &h_fine[16*n*(16*c+k)], &H[c].fine[k][0] );

            #if MEDIAN_HAVE_SIMD
                if( useSIMD )
                {
                    for( j = 0; j < 2*r; ++j )
                        histogram_add_simd( &h_coarse[16*(n*c+j)], H[c].coarse );

                    for( j = r; j < n-r; j++ )
                    {
                        int t = 2*r*r + 2*r, b, sum = 0;
                        HT* segment;

                        histogram_add_simd( &h_coarse[16*(n*c + std::min(j+r,n-1))], H[c].coarse );

                        // Find median at coarse level
                        for ( k = 0; k < 16 ; ++k )
                        {
                            sum += H[c].coarse[k];
                            if ( sum > t )
                            {
                                sum -= H[c].coarse[k];
                                break;
                            }
                        }
                        assert( k < 16 );

                        /* Update corresponding histogram segment */
                        if ( luc[c][k] <= j-r )
                        {
                            memset( &H[c].fine[k], 0, 16 * sizeof(HT) );
                            for ( luc[c][k] = cv::HT(j-r); luc[c][k] < MIN(j+r+1,n); ++luc[c][k] )
                                histogram_add_simd( &h_fine[16*(n*(16*c+k)+luc[c][k])], H[c].fine[k] );

                            if ( luc[c][k] < j+r+1 )
                            {
                                histogram_muladd( j+r+1 - n, &h_fine[16*(n*(16*c+k)+(n-1))], &H[c].fine[k][0] );
                                luc[c][k] = (HT)(j+r+1);
                            }
                        }
                        else
                        {
                            for ( ; luc[c][k] < j+r+1; ++luc[c][k] )
                            {
                                histogram_sub_simd( &h_fine[16*(n*(16*c+k)+MAX(luc[c][k]-2*r-1,0))], H[c].fine[k] );
                                histogram_add_simd( &h_fine[16*(n*(16*c+k)+MIN(luc[c][k],n-1))], H[c].fine[k] );
                            }
                        }

                        histogram_sub_simd( &h_coarse[16*(n*c+MAX(j-r,0))], H[c].coarse );

                        /* Find median in segment */
                        segment = H[c].fine[k];
                        for ( b = 0; b < 16 ; b++ )
                        {
                            sum += segment[b];
                            if ( sum > t )
                            {
                                dst[dstep*i+cn*j+c] = (uchar)(16*k + b);
                                break;
                            }
                        }
                        assert( b < 16 );
                    }
                }
                else
            #endif
                {
                    for( j = 0; j < 2*r; ++j )
                        histogram_add( &h_coarse[16*(n*c+j)], H[c].coarse );

                    for( j = r; j < n-r; j++ )
                    {
                        int t = 2*r*r + 2*r, b, sum = 0;
                        HT* segment;

                        histogram_add( &h_coarse[16*(n*c + std::min(j+r,n-1))], H[c].coarse );

                        // Find median at coarse level
                        for ( k = 0; k < 16 ; ++k )
                        {
                            sum += H[c].coarse[k];
                            if ( sum > t )
                            {
                                sum -= H[c].coarse[k];
                                break;
                            }
                        }
                        assert( k < 16 );

                        /* Update corresponding histogram segment */
                        if ( luc[c][k] <= j-r )
                        {
                            memset( &H[c].fine[k], 0, 16 * sizeof(HT) );
                            for ( luc[c][k] = cv::HT(j-r); luc[c][k] < MIN(j+r+1,n); ++luc[c][k] )
                                histogram_add( &h_fine[16*(n*(16*c+k)+luc[c][k])], H[c].fine[k] );

                            if ( luc[c][k] < j+r+1 )
                            {
                                histogram_muladd( j+r+1 - n, &h_fine[16*(n*(16*c+k)+(n-1))], &H[c].fine[k][0] );
                                luc[c][k] = (HT)(j+r+1);
                            }
                        }
                        else
                        {
                            for ( ; luc[c][k] < j+r+1; ++luc[c][k] )
                            {
                                histogram_sub( &h_fine[16*(n*(16*c+k)+MAX(luc[c][k]-2*r-1,0))], H[c].fine[k] );
                                histogram_add( &h_fine[16*(n*(16*c+k)+MIN(luc[c][k],n-1))], H[c].fine[k] );
                            }
                        }

                        histogram_sub( &h_coarse[16*(n*c+MAX(j-r,0))], H[c].coarse );

                        /* Find median in segment */
                        segment = H[c].fine[k];
                        for ( b = 0; b < 16 ; b++ )
                        {
                            sum += segment[b];
                            if ( sum > t )
                            {
                                dst[dstep*i+cn*j+c] = (uchar)(16*k + b);
                                break;
                            }
                        }
                        assert( b < 16 );
                    }
                }
            }
        }
    }

#undef HOP
#undef COP
}

static void
medianBlur_8u_Om( const Mat& _src, Mat& _dst, int m )
{
    #define N  16
    int     zone0[4][N];
    int     zone1[4][N*N];
    int     x, y;
    int     n2 = m*m/2;
    Size    size = _dst.size();
    const uchar* src = _src.data;
    uchar*  dst = _dst.data;
    int     src_step = (int)_src.step, dst_step = (int)_dst.step;
    int     cn = _src.channels();
    const uchar*  src_max = src + size.height*src_step;

    #define UPDATE_ACC01( pix, cn, op ) \
    {                                   \
        int p = (pix);                  \
        zone1[cn][p] op;                \
        zone0[cn][p >> 4] op;           \
    }

    //CV_Assert( size.height >= nx && size.width >= nx );
    for( x = 0; x < size.width; x++, src += cn, dst += cn )
    {
        uchar* dst_cur = dst;
        const uchar* src_top = src;
        const uchar* src_bottom = src;
        int k, c;
        int src_step1 = src_step, dst_step1 = dst_step;

        if( x % 2 != 0 )
        {
            src_bottom = src_top += src_step*(size.height-1);
            dst_cur += dst_step*(size.height-1);
            src_step1 = -src_step1;
            dst_step1 = -dst_step1;
        }

        // init accumulator
        memset( zone0, 0, sizeof(zone0[0])*cn );
        memset( zone1, 0, sizeof(zone1[0])*cn );

        for( y = 0; y <= m/2; y++ )
        {
            for( c = 0; c < cn; c++ )
            {
                if( y > 0 )
                {
                    for( k = 0; k < m*cn; k += cn )
                        UPDATE_ACC01( src_bottom[k+c], c, ++ );
                }
                else
                {
                    for( k = 0; k < m*cn; k += cn )
                        UPDATE_ACC01( src_bottom[k+c], c, += m/2+1 );
                }
            }

            if( (src_step1 > 0 && y < size.height-1) ||
                (src_step1 < 0 && size.height-y-1 > 0) )
                src_bottom += src_step1;
        }

        for( y = 0; y < size.height; y++, dst_cur += dst_step1 )
        {
            // find median
            for( c = 0; c < cn; c++ )
            {
                int s = 0;
                for( k = 0; ; k++ )
                {
                    int t = s + zone0[c][k];
                    if( t > n2 ) break;
                    s = t;
                }

                for( k *= N; ;k++ )
                {
                    s += zone1[c][k];
                    if( s > n2 ) break;
                }

                dst_cur[c] = (uchar)k;
            }

            if( y+1 == size.height )
                break;

            if( cn == 1 )
            {
                for( k = 0; k < m; k++ )
                {
                    int p = src_top[k];
                    int q = src_bottom[k];
                    zone1[0][p]--;
                    zone0[0][p>>4]--;
                    zone1[0][q]++;
                    zone0[0][q>>4]++;
                }
            }
            else if( cn == 3 )
            {
                for( k = 0; k < m*3; k += 3 )
                {
                    UPDATE_ACC01( src_top[k], 0, -- );
                    UPDATE_ACC01( src_top[k+1], 1, -- );
                    UPDATE_ACC01( src_top[k+2], 2, -- );

                    UPDATE_ACC01( src_bottom[k], 0, ++ );
                    UPDATE_ACC01( src_bottom[k+1], 1, ++ );
                    UPDATE_ACC01( src_bottom[k+2], 2, ++ );
                }
            }
            else
            {
                assert( cn == 4 );
                for( k = 0; k < m*4; k += 4 )
                {
                    UPDATE_ACC01( src_top[k], 0, -- );
                    UPDATE_ACC01( src_top[k+1], 1, -- );
                    UPDATE_ACC01( src_top[k+2], 2, -- );
                    UPDATE_ACC01( src_top[k+3], 3, -- );

                    UPDATE_ACC01( src_bottom[k], 0, ++ );
                    UPDATE_ACC01( src_bottom[k+1], 1, ++ );
                    UPDATE_ACC01( src_bottom[k+2], 2, ++ );
                    UPDATE_ACC01( src_bottom[k+3], 3, ++ );
                }
            }

            if( (src_step1 > 0 && src_bottom + src_step1 < src_max) ||
                (src_step1 < 0 && src_bottom + src_step1 >= src) )
                src_bottom += src_step1;

            if( y >= m/2 )
                src_top += src_step1;
        }
    }
#undef N
#undef UPDATE_ACC
}


struct MinMax8u
{
    typedef uchar value_type;
    typedef int arg_type;
    enum { SIZE = 1 };
    arg_type load(const uchar* ptr) { return *ptr; }
    void store(uchar* ptr, arg_type val) { *ptr = (uchar)val; }
    void operator()(arg_type& a, arg_type& b) const
    {
        int t = CV_FAST_CAST_8U(a - b);
        b += t; a -= t;
    }
};

struct MinMax16u
{
    typedef ushort value_type;
    typedef int arg_type;
    enum { SIZE = 1 };
    arg_type load(const ushort* ptr) { return *ptr; }
    void store(ushort* ptr, arg_type val) { *ptr = (ushort)val; }
    void operator()(arg_type& a, arg_type& b) const
    {
        arg_type t = a;
        a = std::min(a, b);
        b = std::max(b, t);
    }
};

struct MinMax16s
{
    typedef short value_type;
    typedef int arg_type;
    enum { SIZE = 1 };
    arg_type load(const short* ptr) { return *ptr; }
    void store(short* ptr, arg_type val) { *ptr = (short)val; }
    void operator()(arg_type& a, arg_type& b) const
    {
        arg_type t = a;
        a = std::min(a, b);
        b = std::max(b, t);
    }
};

struct MinMax32f
{
    typedef float value_type;
    typedef float arg_type;
    enum { SIZE = 1 };
    arg_type load(const float* ptr) { return *ptr; }
    void store(float* ptr, arg_type val) { *ptr = val; }
    void operator()(arg_type& a, arg_type& b) const
    {
        arg_type t = a;
        a = std::min(a, b);
        b = std::max(b, t);
    }
};

#if CV_SSE2

struct MinMaxVec8u
{
    typedef uchar value_type;
    typedef __m128i arg_type;
    enum { SIZE = 16 };
    arg_type load(const uchar* ptr) { return _mm_loadu_si128((const __m128i*)ptr); }
    void store(uchar* ptr, arg_type val) { _mm_storeu_si128((__m128i*)ptr, val); }
    void operator()(arg_type& a, arg_type& b) const
    {
        arg_type t = a;
        a = _mm_min_epu8(a, b);
        b = _mm_max_epu8(b, t);
    }
};


struct MinMaxVec16u
{
    typedef ushort value_type;
    typedef __m128i arg_type;
    enum { SIZE = 8 };
    arg_type load(const ushort* ptr) { return _mm_loadu_si128((const __m128i*)ptr); }
    void store(ushort* ptr, arg_type val) { _mm_storeu_si128((__m128i*)ptr, val); }
    void operator()(arg_type& a, arg_type& b) const
    {
        arg_type t = _mm_subs_epu16(a, b);
        a = _mm_subs_epu16(a, t);
        b = _mm_adds_epu16(b, t);
    }
};


struct MinMaxVec16s
{
    typedef short value_type;
    typedef __m128i arg_type;
    enum { SIZE = 8 };
    arg_type load(const short* ptr) { return _mm_loadu_si128((const __m128i*)ptr); }
    void store(short* ptr, arg_type val) { _mm_storeu_si128((__m128i*)ptr, val); }
    void operator()(arg_type& a, arg_type& b) const
    {
        arg_type t = a;
        a = _mm_min_epi16(a, b);
        b = _mm_max_epi16(b, t);
    }
};


struct MinMaxVec32f
{
    typedef float value_type;
    typedef __m128 arg_type;
    enum { SIZE = 4 };
    arg_type load(const float* ptr) { return _mm_loadu_ps(ptr); }
    void store(float* ptr, arg_type val) { _mm_storeu_ps(ptr, val); }
    void operator()(arg_type& a, arg_type& b) const
    {
        arg_type t = a;
        a = _mm_min_ps(a, b);
        b = _mm_max_ps(b, t);
    }
};


#else

typedef MinMax8u MinMaxVec8u;
typedef MinMax16u MinMaxVec16u;
typedef MinMax16s MinMaxVec16s;
typedef MinMax32f MinMaxVec32f;

#endif

template<class Op, class VecOp>
static void
medianBlur_SortNet( const Mat& _src, Mat& _dst, int m )
{
    typedef typename Op::value_type T;
    typedef typename Op::arg_type WT;
    typedef typename VecOp::arg_type VT;

    const T* src = (const T*)_src.data;
    T* dst = (T*)_dst.data;
    int sstep = (int)(_src.step/sizeof(T));
    int dstep = (int)(_dst.step/sizeof(T));
    Size size = _dst.size();
    int i, j, k, cn = _src.channels();
    Op op;
    VecOp vop;
    volatile bool useSIMD = checkHardwareSupport(CV_CPU_SSE2);

    if( m == 3 )
    {
        if( size.width == 1 || size.height == 1 )
        {
            int len = size.width + size.height - 1;
            int sdelta = size.height == 1 ? cn : sstep;
            int sdelta0 = size.height == 1 ? 0 : sstep - cn;
            int ddelta = size.height == 1 ? cn : dstep;

            for( i = 0; i < len; i++, src += sdelta0, dst += ddelta )
                for( j = 0; j < cn; j++, src++ )
                {
                    WT p0 = src[i > 0 ? -sdelta : 0];
                    WT p1 = src[0];
                    WT p2 = src[i < len - 1 ? sdelta : 0];

                    op(p0, p1); op(p1, p2); op(p0, p1);
                    dst[j] = (T)p1;
                }
            return;
        }

        size.width *= cn;
        for( i = 0; i < size.height; i++, dst += dstep )
        {
            const T* row0 = src + std::max(i - 1, 0)*sstep;
            const T* row1 = src + i*sstep;
            const T* row2 = src + std::min(i + 1, size.height-1)*sstep;
            int limit = useSIMD ? cn : size.width;

            for(j = 0;; )
            {
                for( ; j < limit; j++ )
                {
                    int j0 = j >= cn ? j - cn : j;
                    int j2 = j < size.width - cn ? j + cn : j;
                    WT p0 = row0[j0], p1 = row0[j], p2 = row0[j2];
                    WT p3 = row1[j0], p4 = row1[j], p5 = row1[j2];
                    WT p6 = row2[j0], p7 = row2[j], p8 = row2[j2];

                    op(p1, p2); op(p4, p5); op(p7, p8); op(p0, p1);
                    op(p3, p4); op(p6, p7); op(p1, p2); op(p4, p5);
                    op(p7, p8); op(p0, p3); op(p5, p8); op(p4, p7);
                    op(p3, p6); op(p1, p4); op(p2, p5); op(p4, p7);
                    op(p4, p2); op(p6, p4); op(p4, p2);
                    dst[j] = (T)p4;
                }

                if( limit == size.width )
                    break;

                for( ; j <= size.width - VecOp::SIZE - cn; j += VecOp::SIZE )
                {
                    VT p0 = vop.load(row0+j-cn), p1 = vop.load(row0+j), p2 = vop.load(row0+j+cn);
                    VT p3 = vop.load(row1+j-cn), p4 = vop.load(row1+j), p5 = vop.load(row1+j+cn);
                    VT p6 = vop.load(row2+j-cn), p7 = vop.load(row2+j), p8 = vop.load(row2+j+cn);

                    vop(p1, p2); vop(p4, p5); vop(p7, p8); vop(p0, p1);
                    vop(p3, p4); vop(p6, p7); vop(p1, p2); vop(p4, p5);
                    vop(p7, p8); vop(p0, p3); vop(p5, p8); vop(p4, p7);
                    vop(p3, p6); vop(p1, p4); vop(p2, p5); vop(p4, p7);
                    vop(p4, p2); vop(p6, p4); vop(p4, p2);
                    vop.store(dst+j, p4);
                }

                limit = size.width;
            }
        }
    }
    else if( m == 5 )
    {
        if( size.width == 1 || size.height == 1 )
        {
            int len = size.width + size.height - 1;
            int sdelta = size.height == 1 ? cn : sstep;
            int sdelta0 = size.height == 1 ? 0 : sstep - cn;
            int ddelta = size.height == 1 ? cn : dstep;

            for( i = 0; i < len; i++, src += sdelta0, dst += ddelta )
                for( j = 0; j < cn; j++, src++ )
                {
                    int i1 = i > 0 ? -sdelta : 0;
                    int i0 = i > 1 ? -sdelta*2 : i1;
                    int i3 = i < len-1 ? sdelta : 0;
                    int i4 = i < len-2 ? sdelta*2 : i3;
                    WT p0 = src[i0], p1 = src[i1], p2 = src[0], p3 = src[i3], p4 = src[i4];

                    op(p0, p1); op(p3, p4); op(p2, p3); op(p3, p4); op(p0, p2);
                    op(p2, p4); op(p1, p3); op(p1, p2);
                    dst[j] = (T)p2;
                }
            return;
        }

        size.width *= cn;
        for( i = 0; i < size.height; i++, dst += dstep )
        {
            const T* row[5];
            row[0] = src + std::max(i - 2, 0)*sstep;
            row[1] = src + std::max(i - 1, 0)*sstep;
            row[2] = src + i*sstep;
            row[3] = src + std::min(i + 1, size.height-1)*sstep;
            row[4] = src + std::min(i + 2, size.height-1)*sstep;
            int limit = useSIMD ? cn*2 : size.width;

            for(j = 0;; )
            {
                for( ; j < limit; j++ )
                {
                    WT p[25];
                    int j1 = j >= cn ? j - cn : j;
                    int j0 = j >= cn*2 ? j - cn*2 : j1;
                    int j3 = j < size.width - cn ? j + cn : j;
                    int j4 = j < size.width - cn*2 ? j + cn*2 : j3;
                    for( k = 0; k < 5; k++ )
                    {
                        const T* rowk = row[k];
                        p[k*5] = rowk[j0]; p[k*5+1] = rowk[j1];
                        p[k*5+2] = rowk[j]; p[k*5+3] = rowk[j3];
                        p[k*5+4] = rowk[j4];
                    }

                    op(p[1], p[2]); op(p[0], p[1]); op(p[1], p[2]); op(p[4], p[5]); op(p[3], p[4]);
                    op(p[4], p[5]); op(p[0], p[3]); op(p[2], p[5]); op(p[2], p[3]); op(p[1], p[4]);
                    op(p[1], p[2]); op(p[3], p[4]); op(p[7], p[8]); op(p[6], p[7]); op(p[7], p[8]);
                    op(p[10], p[11]); op(p[9], p[10]); op(p[10], p[11]); op(p[6], p[9]); op(p[8], p[11]);
                    op(p[8], p[9]); op(p[7], p[10]); op(p[7], p[8]); op(p[9], p[10]); op(p[0], p[6]);
                    op(p[4], p[10]); op(p[4], p[6]); op(p[2], p[8]); op(p[2], p[4]); op(p[6], p[8]);
                    op(p[1], p[7]); op(p[5], p[11]); op(p[5], p[7]); op(p[3], p[9]); op(p[3], p[5]);
                    op(p[7], p[9]); op(p[1], p[2]); op(p[3], p[4]); op(p[5], p[6]); op(p[7], p[8]);
                    op(p[9], p[10]); op(p[13], p[14]); op(p[12], p[13]); op(p[13], p[14]); op(p[16], p[17]);
                    op(p[15], p[16]); op(p[16], p[17]); op(p[12], p[15]); op(p[14], p[17]); op(p[14], p[15]);
                    op(p[13], p[16]); op(p[13], p[14]); op(p[15], p[16]); op(p[19], p[20]); op(p[18], p[19]);
                    op(p[19], p[20]); op(p[21], p[22]); op(p[23], p[24]); op(p[21], p[23]); op(p[22], p[24]);
                    op(p[22], p[23]); op(p[18], p[21]); op(p[20], p[23]); op(p[20], p[21]); op(p[19], p[22]);
                    op(p[22], p[24]); op(p[19], p[20]); op(p[21], p[22]); op(p[23], p[24]); op(p[12], p[18]);
                    op(p[16], p[22]); op(p[16], p[18]); op(p[14], p[20]); op(p[20], p[24]); op(p[14], p[16]);
                    op(p[18], p[20]); op(p[22], p[24]); op(p[13], p[19]); op(p[17], p[23]); op(p[17], p[19]);
                    op(p[15], p[21]); op(p[15], p[17]); op(p[19], p[21]); op(p[13], p[14]); op(p[15], p[16]);
                    op(p[17], p[18]); op(p[19], p[20]); op(p[21], p[22]); op(p[23], p[24]); op(p[0], p[12]);
                    op(p[8], p[20]); op(p[8], p[12]); op(p[4], p[16]); op(p[16], p[24]); op(p[12], p[16]);
                    op(p[2], p[14]); op(p[10], p[22]); op(p[10], p[14]); op(p[6], p[18]); op(p[6], p[10]);
                    op(p[10], p[12]); op(p[1], p[13]); op(p[9], p[21]); op(p[9], p[13]); op(p[5], p[17]);
                    op(p[13], p[17]); op(p[3], p[15]); op(p[11], p[23]); op(p[11], p[15]); op(p[7], p[19]);
                    op(p[7], p[11]); op(p[11], p[13]); op(p[11], p[12]);
                    dst[j] = (T)p[12];
                }

                if( limit == size.width )
                    break;

                for( ; j <= size.width - VecOp::SIZE - cn*2; j += VecOp::SIZE )
                {
                    VT p[25];
                    for( k = 0; k < 5; k++ )
                    {
                        const T* rowk = row[k];
                        p[k*5] = vop.load(rowk+j-cn*2); p[k*5+1] = vop.load(rowk+j-cn);
                        p[k*5+2] = vop.load(rowk+j); p[k*5+3] = vop.load(rowk+j+cn);
                        p[k*5+4] = vop.load(rowk+j+cn*2);
                    }

                    vop(p[1], p[2]); vop(p[0], p[1]); vop(p[1], p[2]); vop(p[4], p[5]); vop(p[3], p[4]);
                    vop(p[4], p[5]); vop(p[0], p[3]); vop(p[2], p[5]); vop(p[2], p[3]); vop(p[1], p[4]);
                    vop(p[1], p[2]); vop(p[3], p[4]); vop(p[7], p[8]); vop(p[6], p[7]); vop(p[7], p[8]);
                    vop(p[10], p[11]); vop(p[9], p[10]); vop(p[10], p[11]); vop(p[6], p[9]); vop(p[8], p[11]);
                    vop(p[8], p[9]); vop(p[7], p[10]); vop(p[7], p[8]); vop(p[9], p[10]); vop(p[0], p[6]);
                    vop(p[4], p[10]); vop(p[4], p[6]); vop(p[2], p[8]); vop(p[2], p[4]); vop(p[6], p[8]);
                    vop(p[1], p[7]); vop(p[5], p[11]); vop(p[5], p[7]); vop(p[3], p[9]); vop(p[3], p[5]);
                    vop(p[7], p[9]); vop(p[1], p[2]); vop(p[3], p[4]); vop(p[5], p[6]); vop(p[7], p[8]);
                    vop(p[9], p[10]); vop(p[13], p[14]); vop(p[12], p[13]); vop(p[13], p[14]); vop(p[16], p[17]);
                    vop(p[15], p[16]); vop(p[16], p[17]); vop(p[12], p[15]); vop(p[14], p[17]); vop(p[14], p[15]);
                    vop(p[13], p[16]); vop(p[13], p[14]); vop(p[15], p[16]); vop(p[19], p[20]); vop(p[18], p[19]);
                    vop(p[19], p[20]); vop(p[21], p[22]); vop(p[23], p[24]); vop(p[21], p[23]); vop(p[22], p[24]);
                    vop(p[22], p[23]); vop(p[18], p[21]); vop(p[20], p[23]); vop(p[20], p[21]); vop(p[19], p[22]);
                    vop(p[22], p[24]); vop(p[19], p[20]); vop(p[21], p[22]); vop(p[23], p[24]); vop(p[12], p[18]);
                    vop(p[16], p[22]); vop(p[16], p[18]); vop(p[14], p[20]); vop(p[20], p[24]); vop(p[14], p[16]);
                    vop(p[18], p[20]); vop(p[22], p[24]); vop(p[13], p[19]); vop(p[17], p[23]); vop(p[17], p[19]);
                    vop(p[15], p[21]); vop(p[15], p[17]); vop(p[19], p[21]); vop(p[13], p[14]); vop(p[15], p[16]);
                    vop(p[17], p[18]); vop(p[19], p[20]); vop(p[21], p[22]); vop(p[23], p[24]); vop(p[0], p[12]);
                    vop(p[8], p[20]); vop(p[8], p[12]); vop(p[4], p[16]); vop(p[16], p[24]); vop(p[12], p[16]);
                    vop(p[2], p[14]); vop(p[10], p[22]); vop(p[10], p[14]); vop(p[6], p[18]); vop(p[6], p[10]);
                    vop(p[10], p[12]); vop(p[1], p[13]); vop(p[9], p[21]); vop(p[9], p[13]); vop(p[5], p[17]);
                    vop(p[13], p[17]); vop(p[3], p[15]); vop(p[11], p[23]); vop(p[11], p[15]); vop(p[7], p[19]);
                    vop(p[7], p[11]); vop(p[11], p[13]); vop(p[11], p[12]);
                    vop.store(dst+j, p[12]);
                }

                limit = size.width;
            }
        }
    }
}

#ifdef HAVE_OPENCL

static bool ocl_medianFilter(InputArray _src, OutputArray _dst, int m)
{
    size_t localsize[2] = { 16, 16 };
    size_t globalsize[2];
    int type = _src.type(), depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);

    if ( !((depth == CV_8U || depth == CV_16U || depth == CV_16S || depth == CV_32F) && cn <= 4 && (m == 3 || m == 5)) )
        return false;

    Size imgSize = _src.size();
    bool useOptimized = (1 == cn) &&
                        (size_t)imgSize.width >= localsize[0] * 8  &&
                        (size_t)imgSize.height >= localsize[1] * 8 &&
                        imgSize.width % 4 == 0 &&
                        imgSize.height % 4 == 0 &&
                        (ocl::Device::getDefault().isIntel());

    cv::String kname = format( useOptimized ? "medianFilter%d_u" : "medianFilter%d", m) ;
    cv::String kdefs = useOptimized ?
                         format("-D T=%s -D T1=%s -D T4=%s%d -D cn=%d -D USE_4OPT", ocl::typeToStr(type),
                         ocl::typeToStr(depth), ocl::typeToStr(depth), cn*4, cn)
                         :
                         format("-D T=%s -D T1=%s -D cn=%d", ocl::typeToStr(type), ocl::typeToStr(depth), cn) ;

    ocl::Kernel k(kname.c_str(), ocl::imgproc::medianFilter_oclsrc, kdefs.c_str() );

    if (k.empty())
        return false;

    UMat src = _src.getUMat();
    _dst.create(src.size(), type);
    UMat dst = _dst.getUMat();

    k.args(ocl::KernelArg::ReadOnlyNoSize(src), ocl::KernelArg::WriteOnly(dst));

    if( useOptimized )
    {
        globalsize[0] = DIVUP(src.cols / 4, localsize[0]) * localsize[0];
        globalsize[1] = DIVUP(src.rows / 4, localsize[1]) * localsize[1];
    }
    else
    {
        globalsize[0] = (src.cols + localsize[0] + 2) / localsize[0] * localsize[0];
        globalsize[1] = (src.rows + localsize[1] - 1) / localsize[1] * localsize[1];
    }

    return k.run(2, globalsize, localsize, false);
}

#endif

}

void cv::medianBlur( InputArray _src0, OutputArray _dst, int ksize )
{
    CV_Assert( (ksize % 2 == 1) && (_src0.dims() <= 2 ));

    if( ksize <= 1 )
    {
        _src0.copyTo(_dst);
        return;
    }

    CV_OCL_RUN(_dst.isUMat(),
               ocl_medianFilter(_src0,_dst, ksize))

    Mat src0 = _src0.getMat();
    _dst.create( src0.size(), src0.type() );
    Mat dst = _dst.getMat();

#if IPP_VERSION_X100 >= 801
#define IPP_FILTER_MEDIAN_BORDER(ippType, ippDataType, flavor) \
    do \
    { \
        if (ippiFilterMedianBorderGetBufferSize(dstRoiSize, maskSize, \
            ippDataType, CV_MAT_CN(type), &bufSize) >= 0) \
        { \
            Ipp8u * buffer = ippsMalloc_8u(bufSize); \
            IppStatus status = ippiFilterMedianBorder_##flavor((const ippType *)src0.data, (int)src0.step, \
                (ippType *)dst.data, (int)dst.step, dstRoiSize, maskSize, \
                ippBorderRepl, (ippType)0, buffer); \
            ippsFree(buffer); \
            if (status >= 0) \
                return; \
        } \
        setIppErrorStatus(); \
    } \
    while ((void)0, 0)

    Ipp32s bufSize;
    IppiSize dstRoiSize = ippiSize(dst.cols, dst.rows), maskSize = ippiSize(ksize, ksize);

    int type = src0.type();
    if (type == CV_8UC1)
        IPP_FILTER_MEDIAN_BORDER(Ipp8u, ipp8u, 8u_C1R);
    else if (type == CV_16UC1)
        IPP_FILTER_MEDIAN_BORDER(Ipp16u, ipp16u, 16u_C1R);
    else if (type == CV_16SC1)
        IPP_FILTER_MEDIAN_BORDER(Ipp16s, ipp16s, 16s_C1R);
    else if (type == CV_32FC1)
        IPP_FILTER_MEDIAN_BORDER(Ipp32f, ipp32f, 32f_C1R);
#undef IPP_FILTER_MEDIAN_BORDER
#endif

#ifdef HAVE_TEGRA_OPTIMIZATION
    if (tegra::medianBlur(src0, dst, ksize))
        return;
#endif

    bool useSortNet = ksize == 3 || (ksize == 5
#if !CV_SSE2
            && src0.depth() > CV_8U
#endif
        );

    Mat src;
    if( useSortNet )
    {
        if( dst.data != src0.data )
            src = src0;
        else
            src0.copyTo(src);

        if( src.depth() == CV_8U )
            medianBlur_SortNet<MinMax8u, MinMaxVec8u>( src, dst, ksize );
        else if( src.depth() == CV_16U )
            medianBlur_SortNet<MinMax16u, MinMaxVec16u>( src, dst, ksize );
        else if( src.depth() == CV_16S )
            medianBlur_SortNet<MinMax16s, MinMaxVec16s>( src, dst, ksize );
        else if( src.depth() == CV_32F )
            medianBlur_SortNet<MinMax32f, MinMaxVec32f>( src, dst, ksize );
        else
            CV_Error(CV_StsUnsupportedFormat, "");

        return;
    }
    else
    {
        cv::copyMakeBorder( src0, src, 0, 0, ksize/2, ksize/2, BORDER_REPLICATE );

        int cn = src0.channels();
        CV_Assert( src.depth() == CV_8U && (cn == 1 || cn == 3 || cn == 4) );

        double img_size_mp = (double)(src0.total())/(1 << 20);
        if( ksize <= 3 + (img_size_mp < 1 ? 12 : img_size_mp < 4 ? 6 : 2)*(MEDIAN_HAVE_SIMD && checkHardwareSupport(CV_CPU_SSE2) ? 1 : 3))
            medianBlur_8u_Om( src, dst, ksize );
        else
            medianBlur_8u_O1( src, dst, ksize );
    }
}

/****************************************************************************************\
                                   Bilateral Filtering
\****************************************************************************************/

namespace cv
{

class BilateralFilter_8u_Invoker :
    public ParallelLoopBody
{
public:
    BilateralFilter_8u_Invoker(Mat& _dest, const Mat& _temp, int _radius, int _maxk,
        int* _space_ofs, float *_space_weight, float *_color_weight) :
        temp(&_temp), dest(&_dest), radius(_radius),
        maxk(_maxk), space_ofs(_space_ofs), space_weight(_space_weight), color_weight(_color_weight)
    {
    }

    virtual void operator() (const Range& range) const
    {
        int i, j, cn = dest->channels(), k;
        Size size = dest->size();
        #if CV_SSE3
        int CV_DECL_ALIGNED(16) buf[4];
        float CV_DECL_ALIGNED(16) bufSum[4];
        static const int CV_DECL_ALIGNED(16) bufSignMask[] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 };
        bool haveSSE3 = checkHardwareSupport(CV_CPU_SSE3);
        #endif

        for( i = range.start; i < range.end; i++ )
        {
            const uchar* sptr = temp->ptr(i+radius) + radius*cn;
            uchar* dptr = dest->ptr(i);

            if( cn == 1 )
            {
                for( j = 0; j < size.width; j++ )
                {
                    float sum = 0, wsum = 0;
                    int val0 = sptr[j];
                    k = 0;
                    #if CV_SSE3
                    if( haveSSE3 )
                    {
                        __m128 _val0 = _mm_set1_ps(static_cast<float>(val0));
                        const __m128 _signMask = _mm_load_ps((const float*)bufSignMask);

                        for( ; k <= maxk - 4; k += 4 )
                        {
                            __m128 _valF = _mm_set_ps(sptr[j + space_ofs[k+3]], sptr[j + space_ofs[k+2]],
                                                      sptr[j + space_ofs[k+1]], sptr[j + space_ofs[k]]);

                            __m128 _val = _mm_andnot_ps(_signMask, _mm_sub_ps(_valF, _val0));
                            _mm_store_si128((__m128i*)buf, _mm_cvtps_epi32(_val));

                            __m128 _cw = _mm_set_ps(color_weight[buf[3]],color_weight[buf[2]],
                                                    color_weight[buf[1]],color_weight[buf[0]]);
                            __m128 _sw = _mm_loadu_ps(space_weight+k);
                            __m128 _w = _mm_mul_ps(_cw, _sw);
                             _cw = _mm_mul_ps(_w, _valF);

                             _sw = _mm_hadd_ps(_w, _cw);
                             _sw = _mm_hadd_ps(_sw, _sw);
                             _mm_storel_pi((__m64*)bufSum, _sw);

                             sum += bufSum[1];
                             wsum += bufSum[0];
                        }
                    }
                    #endif
                    for( ; k < maxk; k++ )
                    {
                        int val = sptr[j + space_ofs[k]];
                        float w = space_weight[k]*color_weight[std::abs(val - val0)];
                        sum += val*w;
                        wsum += w;
                    }
                    // overflow is not possible here => there is no need to use cv::saturate_cast
                    dptr[j] = (uchar)cvRound(sum/wsum);
                }
            }
            else
            {
                assert( cn == 3 );
                for( j = 0; j < size.width*3; j += 3 )
                {
                    float sum_b = 0, sum_g = 0, sum_r = 0, wsum = 0;
                    int b0 = sptr[j], g0 = sptr[j+1], r0 = sptr[j+2];
                    k = 0;
                    #if CV_SSE3
                    if( haveSSE3 )
                    {
                        const __m128i izero = _mm_setzero_si128();
                        const __m128 _b0 = _mm_set1_ps(static_cast<float>(b0));
                        const __m128 _g0 = _mm_set1_ps(static_cast<float>(g0));
                        const __m128 _r0 = _mm_set1_ps(static_cast<float>(r0));
                        const __m128 _signMask = _mm_load_ps((const float*)bufSignMask);

                        for( ; k <= maxk - 4; k += 4 )
                        {
                            const int* const sptr_k0  = reinterpret_cast<const int*>(sptr + j + space_ofs[k]);
                            const int* const sptr_k1  = reinterpret_cast<const int*>(sptr + j + space_ofs[k+1]);
                            const int* const sptr_k2  = reinterpret_cast<const int*>(sptr + j + space_ofs[k+2]);
                            const int* const sptr_k3  = reinterpret_cast<const int*>(sptr + j + space_ofs[k+3]);

                            __m128 _b = _mm_cvtepi32_ps(_mm_unpacklo_epi16(_mm_unpacklo_epi8(_mm_cvtsi32_si128(sptr_k0[0]), izero), izero));
                            __m128 _g = _mm_cvtepi32_ps(_mm_unpacklo_epi16(_mm_unpacklo_epi8(_mm_cvtsi32_si128(sptr_k1[0]), izero), izero));
                            __m128 _r = _mm_cvtepi32_ps(_mm_unpacklo_epi16(_mm_unpacklo_epi8(_mm_cvtsi32_si128(sptr_k2[0]), izero), izero));
                            __m128 _z = _mm_cvtepi32_ps(_mm_unpacklo_epi16(_mm_unpacklo_epi8(_mm_cvtsi32_si128(sptr_k3[0]), izero), izero));

                            _MM_TRANSPOSE4_PS(_b, _g, _r, _z);

                            __m128 bt = _mm_andnot_ps(_signMask, _mm_sub_ps(_b,_b0));
                            __m128 gt = _mm_andnot_ps(_signMask, _mm_sub_ps(_g,_g0));
                            __m128 rt = _mm_andnot_ps(_signMask, _mm_sub_ps(_r,_r0));

                            bt =_mm_add_ps(rt, _mm_add_ps(bt, gt));
                            _mm_store_si128((__m128i*)buf, _mm_cvtps_epi32(bt));

                            __m128 _w  = _mm_set_ps(color_weight[buf[3]],color_weight[buf[2]],
                                                    color_weight[buf[1]],color_weight[buf[0]]);
                            __m128 _sw = _mm_loadu_ps(space_weight+k);

                            _w = _mm_mul_ps(_w,_sw);
                            _b = _mm_mul_ps(_b, _w);
                            _g = _mm_mul_ps(_g, _w);
                            _r = _mm_mul_ps(_r, _w);

                             _w = _mm_hadd_ps(_w, _b);
                             _g = _mm_hadd_ps(_g, _r);

                             _w = _mm_hadd_ps(_w, _g);
                             _mm_store_ps(bufSum, _w);

                             wsum  += bufSum[0];
                             sum_b += bufSum[1];
                             sum_g += bufSum[2];
                             sum_r += bufSum[3];
                         }
                    }
                    #endif

                    for( ; k < maxk; k++ )
                    {
                        const uchar* sptr_k = sptr + j + space_ofs[k];
                        int b = sptr_k[0], g = sptr_k[1], r = sptr_k[2];
                        float w = space_weight[k]*color_weight[std::abs(b - b0) +
                                                               std::abs(g - g0) + std::abs(r - r0)];
                        sum_b += b*w; sum_g += g*w; sum_r += r*w;
                        wsum += w;
                    }
                    wsum = 1.f/wsum;
                    b0 = cvRound(sum_b*wsum);
                    g0 = cvRound(sum_g*wsum);
                    r0 = cvRound(sum_r*wsum);
                    dptr[j] = (uchar)b0; dptr[j+1] = (uchar)g0; dptr[j+2] = (uchar)r0;
                }
            }
        }
    }

private:
    const Mat *temp;
    Mat *dest;
    int radius, maxk, *space_ofs;
    float *space_weight, *color_weight;
};

#if defined (HAVE_IPP) && !defined(HAVE_IPP_ICV_ONLY) && 0
class IPPBilateralFilter_8u_Invoker :
    public ParallelLoopBody
{
public:
    IPPBilateralFilter_8u_Invoker(Mat &_src, Mat &_dst, double _sigma_color, double _sigma_space, int _radius, bool *_ok) :
      ParallelLoopBody(), src(_src), dst(_dst), sigma_color(_sigma_color), sigma_space(_sigma_space), radius(_radius), ok(_ok)
      {
          *ok = true;
      }

      virtual void operator() (const Range& range) const
      {
          int d = radius * 2 + 1;
          IppiSize kernel = {d, d};
          IppiSize roi={dst.cols, range.end - range.start};
          int bufsize=0;
          if (0 > ippiFilterBilateralGetBufSize_8u_C1R( ippiFilterBilateralGauss, roi, kernel, &bufsize))
          {
              *ok = false;
              return;
          }
          AutoBuffer<uchar> buf(bufsize);
          IppiFilterBilateralSpec *pSpec = (IppiFilterBilateralSpec *)alignPtr(&buf[0], 32);
          if (0 > ippiFilterBilateralInit_8u_C1R( ippiFilterBilateralGauss, kernel, (Ipp32f)sigma_color, (Ipp32f)sigma_space, 1, pSpec ))
          {
              *ok = false;
              return;
          }
          if (0 > ippiFilterBilateral_8u_C1R( src.ptr<uchar>(range.start) + radius * ((int)src.step[0] + 1), (int)src.step[0], dst.ptr<uchar>(range.start), (int)dst.step[0], roi, kernel, pSpec ))
              *ok = false;
      }
private:
    Mat &src;
    Mat &dst;
    double sigma_color;
    double sigma_space;
    int radius;
    bool *ok;
    const IPPBilateralFilter_8u_Invoker& operator= (const IPPBilateralFilter_8u_Invoker&);
};
#endif

#ifdef HAVE_OPENCL

static bool ocl_bilateralFilter_8u(InputArray _src, OutputArray _dst, int d,
                                   double sigma_color, double sigma_space,
                                   int borderType)
{
    int type = _src.type(), depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);
    int i, j, maxk, radius;

    if (depth != CV_8U || cn > 4)
        return false;

    if (sigma_color <= 0)
        sigma_color = 1;
    if (sigma_space <= 0)
        sigma_space = 1;

    double gauss_color_coeff = -0.5 / (sigma_color * sigma_color);
    double gauss_space_coeff = -0.5 / (sigma_space * sigma_space);

    if ( d <= 0 )
        radius = cvRound(sigma_space * 1.5);
    else
        radius = d / 2;
    radius = MAX(radius, 1);
    d = radius * 2 + 1;

    UMat src = _src.getUMat(), dst = _dst.getUMat(), temp;
    if (src.u == dst.u)
        return false;

    copyMakeBorder(src, temp, radius, radius, radius, radius, borderType);
    std::vector<float> _space_weight(d * d);
    std::vector<int> _space_ofs(d * d);
    float * const space_weight = &_space_weight[0];
    int * const space_ofs = &_space_ofs[0];

    // initialize space-related bilateral filter coefficients
    for( i = -radius, maxk = 0; i <= radius; i++ )
        for( j = -radius; j <= radius; j++ )
        {
            double r = std::sqrt((double)i * i + (double)j * j);
            if ( r > radius )
                continue;
            space_weight[maxk] = (float)std::exp(r * r * gauss_space_coeff);
            space_ofs[maxk++] = (int)(i * temp.step + j * cn);
        }

    char cvt[3][40];
    String cnstr = cn > 1 ? format("%d", cn) : "";
    String kernelName("bilateral");
    size_t sizeDiv = 1;
    if ((ocl::Device::getDefault().isIntel()) &&
        (ocl::Device::getDefault().type() == ocl::Device::TYPE_GPU))
    {
            //Intel GPU
            if (dst.cols % 4 == 0 && cn == 1) // For single channel x4 sized images.
            {
                kernelName = "bilateral_float4";
                sizeDiv = 4;
            }
     }
     ocl::Kernel k(kernelName.c_str(), ocl::imgproc::bilateral_oclsrc,
            format("-D radius=%d -D maxk=%d -D cn=%d -D int_t=%s -D uint_t=uint%s -D convert_int_t=%s"
            " -D uchar_t=%s -D float_t=%s -D convert_float_t=%s -D convert_uchar_t=%s -D gauss_color_coeff=%f",
            radius, maxk, cn, ocl::typeToStr(CV_32SC(cn)), cnstr.c_str(),
            ocl::convertTypeStr(CV_8U, CV_32S, cn, cvt[0]),
            ocl::typeToStr(type), ocl::typeToStr(CV_32FC(cn)),
            ocl::convertTypeStr(CV_32S, CV_32F, cn, cvt[1]),
            ocl::convertTypeStr(CV_32F, CV_8U, cn, cvt[2]), gauss_color_coeff));
    if (k.empty())
        return false;

    Mat mspace_weight(1, d * d, CV_32FC1, space_weight);
    Mat mspace_ofs(1, d * d, CV_32SC1, space_ofs);
    UMat ucolor_weight, uspace_weight, uspace_ofs;

    mspace_weight.copyTo(uspace_weight);
    mspace_ofs.copyTo(uspace_ofs);

    k.args(ocl::KernelArg::ReadOnlyNoSize(temp), ocl::KernelArg::WriteOnly(dst),
           ocl::KernelArg::PtrReadOnly(uspace_weight),
           ocl::KernelArg::PtrReadOnly(uspace_ofs));

    size_t globalsize[2] = { dst.cols / sizeDiv, dst.rows };
    return k.run(2, globalsize, NULL, false);
}

#endif
static void
bilateralFilter_8u( const Mat& src, Mat& dst, int d,
    double sigma_color, double sigma_space,
    int borderType )
{
    int cn = src.channels();
    int i, j, maxk, radius;
    Size size = src.size();

    CV_Assert( (src.type() == CV_8UC1 || src.type() == CV_8UC3) && src.data != dst.data );

    if( sigma_color <= 0 )
        sigma_color = 1;
    if( sigma_space <= 0 )
        sigma_space = 1;

    double gauss_color_coeff = -0.5/(sigma_color*sigma_color);
    double gauss_space_coeff = -0.5/(sigma_space*sigma_space);

    if( d <= 0 )
        radius = cvRound(sigma_space*1.5);
    else
        radius = d/2;
    radius = MAX(radius, 1);
    d = radius*2 + 1;

    Mat temp;
    copyMakeBorder( src, temp, radius, radius, radius, radius, borderType );

#if defined HAVE_IPP && (IPP_VERSION_MAJOR >= 7) && 0
    if( cn == 1 )
    {
        bool ok;
        IPPBilateralFilter_8u_Invoker body(temp, dst, sigma_color * sigma_color, sigma_space * sigma_space, radius, &ok );
        parallel_for_(Range(0, dst.rows), body, dst.total()/(double)(1<<16));
        if( ok )
            return;
        setIppErrorStatus();
    }
#endif

    std::vector<float> _color_weight(cn*256);
    std::vector<float> _space_weight(d*d);
    std::vector<int> _space_ofs(d*d);
    float* color_weight = &_color_weight[0];
    float* space_weight = &_space_weight[0];
    int* space_ofs = &_space_ofs[0];

    // initialize color-related bilateral filter coefficients

    for( i = 0; i < 256*cn; i++ )
        color_weight[i] = (float)std::exp(i*i*gauss_color_coeff);

    // initialize space-related bilateral filter coefficients
    for( i = -radius, maxk = 0; i <= radius; i++ )
    {
        j = -radius;

        for( ; j <= radius; j++ )
        {
            double r = std::sqrt((double)i*i + (double)j*j);
            if( r > radius )
                continue;
            space_weight[maxk] = (float)std::exp(r*r*gauss_space_coeff);
            space_ofs[maxk++] = (int)(i*temp.step + j*cn);
        }
    }

    BilateralFilter_8u_Invoker body(dst, temp, radius, maxk, space_ofs, space_weight, color_weight);
    parallel_for_(Range(0, size.height), body, dst.total()/(double)(1<<16));
}


class BilateralFilter_32f_Invoker :
    public ParallelLoopBody
{
public:

    BilateralFilter_32f_Invoker(int _cn, int _radius, int _maxk, int *_space_ofs,
        const Mat& _temp, Mat& _dest, float _scale_index, float *_space_weight, float *_expLUT) :
        cn(_cn), radius(_radius), maxk(_maxk), space_ofs(_space_ofs),
        temp(&_temp), dest(&_dest), scale_index(_scale_index), space_weight(_space_weight), expLUT(_expLUT)
    {
    }

    virtual void operator() (const Range& range) const
    {
        int i, j, k;
        Size size = dest->size();
        #if CV_SSE3
        int CV_DECL_ALIGNED(16) idxBuf[4];
        float CV_DECL_ALIGNED(16) bufSum32[4];
        static const int CV_DECL_ALIGNED(16) bufSignMask[] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 };
        bool haveSSE3 = checkHardwareSupport(CV_CPU_SSE3);
        #endif

        for( i = range.start; i < range.end; i++ )
        {
            const float* sptr = temp->ptr<float>(i+radius) + radius*cn;
            float* dptr = dest->ptr<float>(i);

            if( cn == 1 )
            {
                for( j = 0; j < size.width; j++ )
                {
                    float sum = 0, wsum = 0;
                    float val0 = sptr[j];
                    k = 0;
                    #if CV_SSE3
                    if( haveSSE3 )
                    {
                        __m128 psum = _mm_setzero_ps();
                        const __m128 _val0 = _mm_set1_ps(sptr[j]);
                        const __m128 _scale_index = _mm_set1_ps(scale_index);
                        const __m128 _signMask = _mm_load_ps((const float*)bufSignMask);

                        for( ; k <= maxk - 4 ; k += 4 )
                        {
                            __m128 _sw    = _mm_loadu_ps(space_weight + k);
                            __m128 _val   = _mm_set_ps(sptr[j + space_ofs[k+3]], sptr[j + space_ofs[k+2]],
                                                       sptr[j + space_ofs[k+1]], sptr[j + space_ofs[k]]);
                            __m128 _alpha = _mm_mul_ps(_mm_andnot_ps( _signMask, _mm_sub_ps(_val,_val0)), _scale_index);

                            __m128i _idx = _mm_cvtps_epi32(_alpha);
                            _mm_store_si128((__m128i*)idxBuf, _idx);
                            _alpha = _mm_sub_ps(_alpha, _mm_cvtepi32_ps(_idx));

                            __m128 _explut  = _mm_set_ps(expLUT[idxBuf[3]], expLUT[idxBuf[2]],
                                                         expLUT[idxBuf[1]], expLUT[idxBuf[0]]);
                            __m128 _explut1 = _mm_set_ps(expLUT[idxBuf[3]+1], expLUT[idxBuf[2]+1],
                                                         expLUT[idxBuf[1]+1], expLUT[idxBuf[0]+1]);

                            __m128 _w = _mm_mul_ps(_sw, _mm_add_ps(_explut, _mm_mul_ps(_alpha, _mm_sub_ps(_explut1, _explut))));
                            _val = _mm_mul_ps(_w, _val);

                             _sw = _mm_hadd_ps(_w, _val);
                             _sw = _mm_hadd_ps(_sw, _sw);
                             psum = _mm_add_ps(_sw, psum);
                        }
                        _mm_storel_pi((__m64*)bufSum32, psum);

                        sum = bufSum32[1];
                        wsum = bufSum32[0];
                    }
                    #endif

                    for( ; k < maxk; k++ )
                    {
                        float val = sptr[j + space_ofs[k]];
                        float alpha = (float)(std::abs(val - val0)*scale_index);
                        int idx = cvFloor(alpha);
                        alpha -= idx;
                        float w = space_weight[k]*(expLUT[idx] + alpha*(expLUT[idx+1] - expLUT[idx]));
                        sum += val*w;
                        wsum += w;
                    }
                    dptr[j] = (float)(sum/wsum);
                }
            }
            else
            {
                CV_Assert( cn == 3 );
                for( j = 0; j < size.width*3; j += 3 )
                {
                    float sum_b = 0, sum_g = 0, sum_r = 0, wsum = 0;
                    float b0 = sptr[j], g0 = sptr[j+1], r0 = sptr[j+2];
                    k = 0;
                    #if  CV_SSE3
                    if( haveSSE3 )
                    {
                        __m128 sum = _mm_setzero_ps();
                        const __m128 _b0 = _mm_set1_ps(b0);
                        const __m128 _g0 = _mm_set1_ps(g0);
                        const __m128 _r0 = _mm_set1_ps(r0);
                        const __m128 _scale_index = _mm_set1_ps(scale_index);
                        const __m128 _signMask = _mm_load_ps((const float*)bufSignMask);

                        for( ; k <= maxk-4; k += 4 )
                        {
                            __m128 _sw = _mm_loadu_ps(space_weight + k);

                            const float* const sptr_k0 = sptr + j + space_ofs[k];
                            const float* const sptr_k1 = sptr + j + space_ofs[k+1];
                            const float* const sptr_k2 = sptr + j + space_ofs[k+2];
                            const float* const sptr_k3 = sptr + j + space_ofs[k+3];

                            __m128 _b = _mm_loadu_ps(sptr_k0);
                            __m128 _g = _mm_loadu_ps(sptr_k1);
                            __m128 _r = _mm_loadu_ps(sptr_k2);
                            __m128 _z = _mm_loadu_ps(sptr_k3);
                            _MM_TRANSPOSE4_PS(_b, _g, _r, _z);

                            __m128 _bt = _mm_andnot_ps(_signMask,_mm_sub_ps(_b,_b0));
                            __m128 _gt = _mm_andnot_ps(_signMask,_mm_sub_ps(_g,_g0));
                            __m128 _rt = _mm_andnot_ps(_signMask,_mm_sub_ps(_r,_r0));

                            __m128 _alpha = _mm_mul_ps(_scale_index, _mm_add_ps(_rt,_mm_add_ps(_bt, _gt)));

                            __m128i _idx  = _mm_cvtps_epi32(_alpha);
                            _mm_store_si128((__m128i*)idxBuf, _idx);
                            _alpha = _mm_sub_ps(_alpha, _mm_cvtepi32_ps(_idx));

                            __m128 _explut  = _mm_set_ps(expLUT[idxBuf[3]], expLUT[idxBuf[2]], expLUT[idxBuf[1]], expLUT[idxBuf[0]]);
                            __m128 _explut1 = _mm_set_ps(expLUT[idxBuf[3]+1], expLUT[idxBuf[2]+1], expLUT[idxBuf[1]+1], expLUT[idxBuf[0]+1]);

                            __m128 _w = _mm_mul_ps(_sw, _mm_add_ps(_explut, _mm_mul_ps(_alpha, _mm_sub_ps(_explut1, _explut))));

                            _b = _mm_mul_ps(_b, _w);
                            _g = _mm_mul_ps(_g, _w);
                            _r = _mm_mul_ps(_r, _w);

                             _w = _mm_hadd_ps(_w, _b);
                             _g = _mm_hadd_ps(_g, _r);

                             _w = _mm_hadd_ps(_w, _g);
                             sum = _mm_add_ps(sum, _w);
                        }
                        _mm_store_ps(bufSum32, sum);
                        wsum  = bufSum32[0];
                        sum_b = bufSum32[1];
                        sum_g = bufSum32[2];
                        sum_r = bufSum32[3];
                    }
                    #endif

                    for(; k < maxk; k++ )
                    {
                        const float* sptr_k = sptr + j + space_ofs[k];
                        float b = sptr_k[0], g = sptr_k[1], r = sptr_k[2];
                        float alpha = (float)((std::abs(b - b0) +
                            std::abs(g - g0) + std::abs(r - r0))*scale_index);
                        int idx = cvFloor(alpha);
                        alpha -= idx;
                        float w = space_weight[k]*(expLUT[idx] + alpha*(expLUT[idx+1] - expLUT[idx]));
                        sum_b += b*w; sum_g += g*w; sum_r += r*w;
                        wsum += w;
                    }
                    wsum = 1.f/wsum;
                    b0 = sum_b*wsum;
                    g0 = sum_g*wsum;
                    r0 = sum_r*wsum;
                    dptr[j] = b0; dptr[j+1] = g0; dptr[j+2] = r0;
                }
            }
        }
    }

private:
    int cn, radius, maxk, *space_ofs;
    const Mat* temp;
    Mat *dest;
    float scale_index, *space_weight, *expLUT;
};


static void
bilateralFilter_32f( const Mat& src, Mat& dst, int d,
                     double sigma_color, double sigma_space,
                     int borderType )
{
    int cn = src.channels();
    int i, j, maxk, radius;
    double minValSrc=-1, maxValSrc=1;
    const int kExpNumBinsPerChannel = 1 << 12;
    int kExpNumBins = 0;
    float lastExpVal = 1.f;
    float len, scale_index;
    Size size = src.size();

    CV_Assert( (src.type() == CV_32FC1 || src.type() == CV_32FC3) && src.data != dst.data );

    if( sigma_color <= 0 )
        sigma_color = 1;
    if( sigma_space <= 0 )
        sigma_space = 1;

    double gauss_color_coeff = -0.5/(sigma_color*sigma_color);
    double gauss_space_coeff = -0.5/(sigma_space*sigma_space);

    if( d <= 0 )
        radius = cvRound(sigma_space*1.5);
    else
        radius = d/2;
    radius = MAX(radius, 1);
    d = radius*2 + 1;
    // compute the min/max range for the input image (even if multichannel)

    minMaxLoc( src.reshape(1), &minValSrc, &maxValSrc );
    if(std::abs(minValSrc - maxValSrc) < FLT_EPSILON)
    {
        src.copyTo(dst);
        return;
    }

    // temporary copy of the image with borders for easy processing
    Mat temp;
    copyMakeBorder( src, temp, radius, radius, radius, radius, borderType );
    const double insteadNaNValue = -5. * sigma_color;
    patchNaNs( temp, insteadNaNValue ); // this replacement of NaNs makes the assumption that depth values are nonnegative
                                        // TODO: make insteadNaNValue avalible in the outside function interface to control the cases breaking the assumption
    // allocate lookup tables
    std::vector<float> _space_weight(d*d);
    std::vector<int> _space_ofs(d*d);
    float* space_weight = &_space_weight[0];
    int* space_ofs = &_space_ofs[0];

    // assign a length which is slightly more than needed
    len = (float)(maxValSrc - minValSrc) * cn;
    kExpNumBins = kExpNumBinsPerChannel * cn;
    std::vector<float> _expLUT(kExpNumBins+2);
    float* expLUT = &_expLUT[0];

    scale_index = kExpNumBins/len;

    // initialize the exp LUT
    for( i = 0; i < kExpNumBins+2; i++ )
    {
        if( lastExpVal > 0.f )
        {
            double val =  i / scale_index;
            expLUT[i] = (float)std::exp(val * val * gauss_color_coeff);
            lastExpVal = expLUT[i];
        }
        else
            expLUT[i] = 0.f;
    }

    // initialize space-related bilateral filter coefficients
    for( i = -radius, maxk = 0; i <= radius; i++ )
        for( j = -radius; j <= radius; j++ )
        {
            double r = std::sqrt((double)i*i + (double)j*j);
            if( r > radius )
                continue;
            space_weight[maxk] = (float)std::exp(r*r*gauss_space_coeff);
            space_ofs[maxk++] = (int)(i*(temp.step/sizeof(float)) + j*cn);
        }

    // parallel_for usage

    BilateralFilter_32f_Invoker body(cn, radius, maxk, space_ofs, temp, dst, scale_index, space_weight, expLUT);
    parallel_for_(Range(0, size.height), body, dst.total()/(double)(1<<16));
}

}

void cv::bilateralFilter( InputArray _src, OutputArray _dst, int d,
                      double sigmaColor, double sigmaSpace,
                      int borderType )
{
    _dst.create( _src.size(), _src.type() );

    CV_OCL_RUN(_src.dims() <= 2 && _dst.isUMat(),
               ocl_bilateralFilter_8u(_src, _dst, d, sigmaColor, sigmaSpace, borderType))

    Mat src = _src.getMat(), dst = _dst.getMat();

    if( src.depth() == CV_8U )
        bilateralFilter_8u( src, dst, d, sigmaColor, sigmaSpace, borderType );
    else if( src.depth() == CV_32F )
        bilateralFilter_32f( src, dst, d, sigmaColor, sigmaSpace, borderType );
    else
        CV_Error( CV_StsUnsupportedFormat,
        "Bilateral filtering is only implemented for 8u and 32f images" );
}

//////////////////////////////////////////////////////////////////////////////////////////

CV_IMPL void
cvSmooth( const void* srcarr, void* dstarr, int smooth_type,
          int param1, int param2, double param3, double param4 )
{
    cv::Mat src = cv::cvarrToMat(srcarr), dst0 = cv::cvarrToMat(dstarr), dst = dst0;

    CV_Assert( dst.size() == src.size() &&
        (smooth_type == CV_BLUR_NO_SCALE || dst.type() == src.type()) );

    if( param2 <= 0 )
        param2 = param1;

    if( smooth_type == CV_BLUR || smooth_type == CV_BLUR_NO_SCALE )
        cv::boxFilter( src, dst, dst.depth(), cv::Size(param1, param2), cv::Point(-1,-1),
            smooth_type == CV_BLUR, cv::BORDER_REPLICATE );
    else if( smooth_type == CV_GAUSSIAN )
        cv::GaussianBlur( src, dst, cv::Size(param1, param2), param3, param4, cv::BORDER_REPLICATE );
    else if( smooth_type == CV_MEDIAN )
        cv::medianBlur( src, dst, param1 );
    else
        cv::bilateralFilter( src, dst, param1, param3, param4, cv::BORDER_REPLICATE );

    if( dst.data != dst0.data )
        CV_Error( CV_StsUnmatchedFormats, "The destination image does not have the proper type" );
}

/* End of file. */