mirror of
https://github.com/opencv/opencv.git
synced 2024-11-25 19:50:38 +08:00
256 lines
9.6 KiB
C++
256 lines
9.6 KiB
C++
/*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 "test_precomp.hpp"
|
|
#include <time.h>
|
|
#include <limits>
|
|
using namespace cv;
|
|
using namespace std;
|
|
|
|
#define CORE_COUNTNONZERO_ERROR_COUNT 1
|
|
|
|
#define MESSAGE_ERROR_COUNT "Count non zero elements returned by OpenCV function is incorrect."
|
|
|
|
#define sign(a) a > 0 ? 1 : a == 0 ? 0 : -1
|
|
|
|
const int FLOAT_TYPE [2] = {CV_32F, CV_64F};
|
|
const int INT_TYPE [5] = {CV_8U, CV_8S, CV_16U, CV_16S, CV_32S};
|
|
|
|
#define MAX_WIDTH 100
|
|
#define MAX_HEIGHT 100
|
|
|
|
class CV_CountNonZeroTest: public cvtest::BaseTest
|
|
{
|
|
public:
|
|
CV_CountNonZeroTest();
|
|
~CV_CountNonZeroTest();
|
|
|
|
protected:
|
|
void run (int);
|
|
|
|
private:
|
|
float eps_32;
|
|
double eps_64;
|
|
Mat src;
|
|
int current_type;
|
|
|
|
void generate_src_data(cv::Size size, int type);
|
|
void generate_src_data(cv::Size size, int type, int count_non_zero);
|
|
void generate_src_stat_data(cv::Size size, int type, int distribution);
|
|
|
|
int get_count_non_zero();
|
|
|
|
void print_information(int right, int result);
|
|
};
|
|
|
|
CV_CountNonZeroTest::CV_CountNonZeroTest(): eps_32(std::numeric_limits<float>::min()), eps_64(std::numeric_limits<double>::min()), src(Mat()), current_type(-1) {}
|
|
CV_CountNonZeroTest::~CV_CountNonZeroTest() {}
|
|
|
|
void CV_CountNonZeroTest::generate_src_data(cv::Size size, int type)
|
|
{
|
|
src.create(size, CV_MAKETYPE(type, 1));
|
|
|
|
for (int j = 0; j < size.width; ++j)
|
|
for (int i = 0; i < size.height; ++i)
|
|
switch (type)
|
|
{
|
|
case CV_8U: { src.at<uchar>(i, j) = cv::randu<uchar>(); break; }
|
|
case CV_8S: { src.at<char>(i, j) = cv::randu<uchar>() - 128; break; }
|
|
case CV_16U: { src.at<ushort>(i, j) = cv::randu<ushort>(); break; }
|
|
case CV_16S: { src.at<short>(i, j) = cv::randu<short>(); break; }
|
|
case CV_32S: { src.at<int>(i, j) = cv::randu<int>(); break; }
|
|
case CV_32F: { src.at<float>(i, j) = cv::randu<float>(); break; }
|
|
case CV_64F: { src.at<double>(i, j) = cv::randu<double>(); break; }
|
|
default: break;
|
|
}
|
|
}
|
|
|
|
void CV_CountNonZeroTest::generate_src_data(cv::Size size, int type, int count_non_zero)
|
|
{
|
|
src = Mat::zeros(size, CV_MAKETYPE(type, 1));
|
|
|
|
int n = 0; RNG& rng = ts->get_rng();
|
|
|
|
while (n < count_non_zero)
|
|
{
|
|
int i = rng.next()%size.height, j = rng.next()%size.width;
|
|
|
|
switch (type)
|
|
{
|
|
case CV_8U: { if (!src.at<uchar>(i, j)) {src.at<uchar>(i, j) = cv::randu<uchar>(); n += (src.at<uchar>(i, j) > 0);} break; }
|
|
case CV_8S: { if (!src.at<char>(i, j)) {src.at<char>(i, j) = cv::randu<uchar>() - 128; n += abs(sign(src.at<char>(i, j)));} break; }
|
|
case CV_16U: { if (!src.at<ushort>(i, j)) {src.at<ushort>(i, j) = cv::randu<ushort>(); n += (src.at<ushort>(i, j) > 0);} break; }
|
|
case CV_16S: { if (!src.at<short>(i, j)) {src.at<short>(i, j) = cv::randu<short>(); n += abs(sign(src.at<short>(i, j)));} break; }
|
|
case CV_32S: { if (!src.at<int>(i, j)) {src.at<int>(i, j) = cv::randu<int>(); n += abs(sign(src.at<int>(i, j)));} break; }
|
|
case CV_32F: { if (fabs(src.at<float>(i, j)) <= eps_32) {src.at<float>(i, j) = cv::randu<float>(); n += (fabs(src.at<float>(i, j)) > eps_32);} break; }
|
|
case CV_64F: { if (fabs(src.at<double>(i, j)) <= eps_64) {src.at<double>(i, j) = cv::randu<double>(); n += (fabs(src.at<double>(i, j)) > eps_64);} break; }
|
|
|
|
default: break;
|
|
}
|
|
}
|
|
|
|
}
|
|
|
|
void CV_CountNonZeroTest::generate_src_stat_data(cv::Size size, int type, int distribution)
|
|
{
|
|
src.create(size, CV_MAKETYPE(type, 1));
|
|
|
|
double mean = 0.0, sigma = 1.0;
|
|
double left = -1.0, right = 1.0;
|
|
|
|
RNG& rng = ts->get_rng();
|
|
|
|
if (distribution == RNG::NORMAL)
|
|
rng.fill(src, RNG::NORMAL, Scalar::all(mean), Scalar::all(sigma));
|
|
else if (distribution == RNG::UNIFORM)
|
|
rng.fill(src, RNG::UNIFORM, Scalar::all(left), Scalar::all(right));
|
|
}
|
|
|
|
int CV_CountNonZeroTest::get_count_non_zero()
|
|
{
|
|
int result = 0;
|
|
|
|
for (int i = 0; i < src.rows; ++i)
|
|
for (int j = 0; j < src.cols; ++j)
|
|
{
|
|
if (current_type == CV_8U) result += (src.at<uchar>(i, j) > 0);
|
|
else if (current_type == CV_8S) result += abs(sign(src.at<char>(i, j)));
|
|
else if (current_type == CV_16U) result += (src.at<ushort>(i, j) > 0);
|
|
else if (current_type == CV_16S) result += abs(sign(src.at<short>(i, j)));
|
|
else if (current_type == CV_32S) result += abs(sign(src.at<int>(i, j)));
|
|
else if (current_type == CV_32F) result += (fabs(src.at<float>(i, j)) > eps_32);
|
|
else result += (fabs(src.at<double>(i, j)) > eps_64);
|
|
}
|
|
|
|
return result;
|
|
}
|
|
|
|
void CV_CountNonZeroTest::print_information(int right, int result)
|
|
{
|
|
cout << endl; cout << "Checking for the work of countNonZero function..." << endl; cout << endl;
|
|
cout << "Type of Mat: ";
|
|
switch (current_type)
|
|
{
|
|
case 0: {cout << "CV_8U"; break;}
|
|
case 1: {cout << "CV_8S"; break;}
|
|
case 2: {cout << "CV_16U"; break;}
|
|
case 3: {cout << "CV_16S"; break;}
|
|
case 4: {cout << "CV_32S"; break;}
|
|
case 5: {cout << "CV_32F"; break;}
|
|
case 6: {cout << "CV_64F"; break;}
|
|
default: break;
|
|
}
|
|
cout << endl;
|
|
cout << "Number of rows: " << src.rows << " Number of cols: " << src.cols << endl;
|
|
cout << "True count non zero elements: " << right << " Result: " << result << endl;
|
|
cout << endl;
|
|
}
|
|
|
|
void CV_CountNonZeroTest::run(int)
|
|
{
|
|
const size_t N = 1500;
|
|
|
|
for (int k = 1; k <= 3; ++k)
|
|
for (size_t i = 0; i < N; ++i)
|
|
{
|
|
RNG& rng = ts->get_rng();
|
|
|
|
int w = rng.next()%MAX_WIDTH + 1, h = rng.next()%MAX_HEIGHT + 1;
|
|
|
|
current_type = rng.next()%7;
|
|
|
|
switch (k)
|
|
{
|
|
case 1: {
|
|
generate_src_data(Size(w, h), current_type);
|
|
int right = get_count_non_zero(), result = countNonZero(src);
|
|
if (result != right)
|
|
{
|
|
cout << "Number of experiment: " << i << endl;
|
|
cout << "Method of data generation: RANDOM" << endl;
|
|
print_information(right, result);
|
|
CV_Error(CORE_COUNTNONZERO_ERROR_COUNT, MESSAGE_ERROR_COUNT);
|
|
return;
|
|
}
|
|
|
|
break;
|
|
}
|
|
|
|
case 2: {
|
|
int count_non_zero = rng.next()%(w*h);
|
|
generate_src_data(Size(w, h), current_type, count_non_zero);
|
|
int result = countNonZero(src);
|
|
if (result != count_non_zero)
|
|
{
|
|
cout << "Number of experiment: " << i << endl;
|
|
cout << "Method of data generation: HALF-RANDOM" << endl;
|
|
print_information(count_non_zero, result);
|
|
CV_Error(CORE_COUNTNONZERO_ERROR_COUNT, MESSAGE_ERROR_COUNT);
|
|
return;
|
|
}
|
|
|
|
break;
|
|
}
|
|
|
|
case 3: {
|
|
int distribution = cv::randu<uchar>()%2;
|
|
generate_src_stat_data(Size(w, h), current_type, distribution);
|
|
int right = get_count_non_zero(), result = countNonZero(src);
|
|
if (right != result)
|
|
{
|
|
cout << "Number of experiment: " << i << endl;
|
|
cout << "Method of data generation: STATISTIC" << endl;
|
|
print_information(right, result);
|
|
CV_Error(CORE_COUNTNONZERO_ERROR_COUNT, MESSAGE_ERROR_COUNT);
|
|
return;
|
|
}
|
|
|
|
break;
|
|
}
|
|
|
|
default: break;
|
|
}
|
|
}
|
|
}
|
|
|
|
TEST (Core_CountNonZero, accuracy) { CV_CountNonZeroTest test; test.safe_run(); }
|