mirror of
https://github.com/tesseract-ocr/tesseract.git
synced 2024-12-11 23:19:04 +08:00
e74d997668
Signed-off-by: Stefan Weil <sw@weilnetz.de>
1020 lines
38 KiB
C++
1020 lines
38 KiB
C++
/* -*-C-*-
|
|
********************************************************************************
|
|
*
|
|
* File: blobs.cpp (Formerly blobs.c)
|
|
* Description: Blob definition
|
|
* Author: Mark Seaman, OCR Technology
|
|
* Created: Fri Oct 27 15:39:52 1989
|
|
* Modified: Thu Mar 28 15:33:26 1991 (Mark Seaman) marks@hpgrlt
|
|
* Language: C
|
|
* Package: N/A
|
|
* Status: Experimental (Do Not Distribute)
|
|
*
|
|
* (c) Copyright 1989, Hewlett-Packard Company.
|
|
** Licensed under the Apache License, Version 2.0 (the "License");
|
|
** you may not use this file except in compliance with the License.
|
|
** You may obtain a copy of the License at
|
|
** http://www.apache.org/licenses/LICENSE-2.0
|
|
** Unless required by applicable law or agreed to in writing, software
|
|
** distributed under the License is distributed on an "AS IS" BASIS,
|
|
** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
** See the License for the specific language governing permissions and
|
|
** limitations under the License.
|
|
*
|
|
*********************************************************************************/
|
|
|
|
/*----------------------------------------------------------------------
|
|
I n c l u d e s
|
|
----------------------------------------------------------------------*/
|
|
// Include automatically generated configuration file if running autoconf.
|
|
#ifdef HAVE_CONFIG_H
|
|
#include "config_auto.h"
|
|
#endif
|
|
|
|
#include "blobs.h"
|
|
#include "ccstruct.h"
|
|
#include "clst.h"
|
|
#include "cutil.h"
|
|
#include "emalloc.h"
|
|
#include "helpers.h"
|
|
#include "linlsq.h"
|
|
#include "ndminx.h"
|
|
#include "normalis.h"
|
|
#include "ocrblock.h"
|
|
#include "ocrrow.h"
|
|
#include "points.h"
|
|
#include "polyaprx.h"
|
|
#include "structures.h"
|
|
#include "werd.h"
|
|
|
|
using tesseract::CCStruct;
|
|
|
|
// A Vector representing the "vertical" direction when measuring the
|
|
// divisiblity of blobs into multiple blobs just by separating outlines.
|
|
// See divisible_blob below for the use.
|
|
const TPOINT kDivisibleVerticalUpright(0, 1);
|
|
// A vector representing the "vertical" direction for italic text for use
|
|
// when separating outlines. Using it actually deteriorates final accuracy,
|
|
// so it is only used for ApplyBoxes chopping to get a better segmentation.
|
|
const TPOINT kDivisibleVerticalItalic(1, 5);
|
|
|
|
/*----------------------------------------------------------------------
|
|
F u n c t i o n s
|
|
----------------------------------------------------------------------*/
|
|
|
|
CLISTIZE(EDGEPT)
|
|
|
|
// Returns true when the two line segments cross each other.
|
|
// (Moved from outlines.cpp).
|
|
// Finds where the projected lines would cross and then checks to see if the
|
|
// point of intersection lies on both of the line segments. If it does
|
|
// then these two segments cross.
|
|
/* static */
|
|
bool TPOINT::IsCrossed(const TPOINT& a0, const TPOINT& a1, const TPOINT& b0,
|
|
const TPOINT& b1) {
|
|
int b0a1xb0b1, b0b1xb0a0;
|
|
int a1b1xa1a0, a1a0xa1b0;
|
|
|
|
TPOINT b0a1, b0a0, a1b1, b0b1, a1a0;
|
|
|
|
b0a1.x = a1.x - b0.x;
|
|
b0a0.x = a0.x - b0.x;
|
|
a1b1.x = b1.x - a1.x;
|
|
b0b1.x = b1.x - b0.x;
|
|
a1a0.x = a0.x - a1.x;
|
|
b0a1.y = a1.y - b0.y;
|
|
b0a0.y = a0.y - b0.y;
|
|
a1b1.y = b1.y - a1.y;
|
|
b0b1.y = b1.y - b0.y;
|
|
a1a0.y = a0.y - a1.y;
|
|
|
|
b0a1xb0b1 = CROSS(b0a1, b0b1);
|
|
b0b1xb0a0 = CROSS(b0b1, b0a0);
|
|
a1b1xa1a0 = CROSS(a1b1, a1a0);
|
|
// For clarity, we want CROSS(a1a0,a1b0) here but we have b0a1 instead of a1b0
|
|
// so use -CROSS(a1b0,b0a1) instead, which is the same.
|
|
a1a0xa1b0 = -CROSS(a1a0, b0a1);
|
|
|
|
return ((b0a1xb0b1 > 0 && b0b1xb0a0 > 0) ||
|
|
(b0a1xb0b1 < 0 && b0b1xb0a0 < 0)) &&
|
|
((a1b1xa1a0 > 0 && a1a0xa1b0 > 0) || (a1b1xa1a0 < 0 && a1a0xa1b0 < 0));
|
|
}
|
|
|
|
// Consume the circular list of EDGEPTs to make a TESSLINE.
|
|
TESSLINE* TESSLINE::BuildFromOutlineList(EDGEPT* outline) {
|
|
TESSLINE* result = new TESSLINE;
|
|
result->loop = outline;
|
|
if (outline->src_outline != nullptr) {
|
|
// ASSUMPTION: This function is only ever called from ApproximateOutline
|
|
// and therefore either all points have a src_outline or all do not.
|
|
// Just as SetupFromPos sets the vectors from the vertices, setup the
|
|
// step_count members to indicate the (positive) number of original
|
|
// C_OUTLINE steps to the next vertex.
|
|
EDGEPT* pt = outline;
|
|
do {
|
|
pt->step_count = pt->next->start_step - pt->start_step;
|
|
if (pt->step_count < 0)
|
|
pt->step_count += pt->src_outline->pathlength();
|
|
pt = pt->next;
|
|
} while (pt != outline);
|
|
}
|
|
result->SetupFromPos();
|
|
return result;
|
|
}
|
|
|
|
// Copies the data and the outline, but leaves next untouched.
|
|
void TESSLINE::CopyFrom(const TESSLINE& src) {
|
|
Clear();
|
|
topleft = src.topleft;
|
|
botright = src.botright;
|
|
start = src.start;
|
|
is_hole = src.is_hole;
|
|
if (src.loop != nullptr) {
|
|
EDGEPT* prevpt = nullptr;
|
|
EDGEPT* newpt = nullptr;
|
|
EDGEPT* srcpt = src.loop;
|
|
do {
|
|
newpt = new EDGEPT(*srcpt);
|
|
if (prevpt == nullptr) {
|
|
loop = newpt;
|
|
} else {
|
|
newpt->prev = prevpt;
|
|
prevpt->next = newpt;
|
|
}
|
|
prevpt = newpt;
|
|
srcpt = srcpt->next;
|
|
} while (srcpt != src.loop);
|
|
loop->prev = newpt;
|
|
newpt->next = loop;
|
|
}
|
|
}
|
|
|
|
// Deletes owned data.
|
|
void TESSLINE::Clear() {
|
|
if (loop == nullptr)
|
|
return;
|
|
|
|
EDGEPT* this_edge = loop;
|
|
do {
|
|
EDGEPT* next_edge = this_edge->next;
|
|
delete this_edge;
|
|
this_edge = next_edge;
|
|
} while (this_edge != loop);
|
|
loop = nullptr;
|
|
}
|
|
|
|
// Normalize in-place using the DENORM.
|
|
void TESSLINE::Normalize(const DENORM& denorm) {
|
|
EDGEPT* pt = loop;
|
|
do {
|
|
denorm.LocalNormTransform(pt->pos, &pt->pos);
|
|
pt = pt->next;
|
|
} while (pt != loop);
|
|
SetupFromPos();
|
|
}
|
|
|
|
// Rotates by the given rotation in place.
|
|
void TESSLINE::Rotate(const FCOORD rot) {
|
|
EDGEPT* pt = loop;
|
|
do {
|
|
int tmp = static_cast<int>(floor(pt->pos.x * rot.x() -
|
|
pt->pos.y * rot.y() + 0.5));
|
|
pt->pos.y = static_cast<int>(floor(pt->pos.y * rot.x() +
|
|
pt->pos.x * rot.y() + 0.5));
|
|
pt->pos.x = tmp;
|
|
pt = pt->next;
|
|
} while (pt != loop);
|
|
SetupFromPos();
|
|
}
|
|
|
|
// Moves by the given vec in place.
|
|
void TESSLINE::Move(const ICOORD vec) {
|
|
EDGEPT* pt = loop;
|
|
do {
|
|
pt->pos.x += vec.x();
|
|
pt->pos.y += vec.y();
|
|
pt = pt->next;
|
|
} while (pt != loop);
|
|
SetupFromPos();
|
|
}
|
|
|
|
// Scales by the given factor in place.
|
|
void TESSLINE::Scale(float factor) {
|
|
EDGEPT* pt = loop;
|
|
do {
|
|
pt->pos.x = static_cast<int>(floor(pt->pos.x * factor + 0.5));
|
|
pt->pos.y = static_cast<int>(floor(pt->pos.y * factor + 0.5));
|
|
pt = pt->next;
|
|
} while (pt != loop);
|
|
SetupFromPos();
|
|
}
|
|
|
|
// Sets up the start and vec members of the loop from the pos members.
|
|
void TESSLINE::SetupFromPos() {
|
|
EDGEPT* pt = loop;
|
|
do {
|
|
pt->vec.x = pt->next->pos.x - pt->pos.x;
|
|
pt->vec.y = pt->next->pos.y - pt->pos.y;
|
|
pt = pt->next;
|
|
} while (pt != loop);
|
|
start = pt->pos;
|
|
ComputeBoundingBox();
|
|
}
|
|
|
|
// Recomputes the bounding box from the points in the loop.
|
|
void TESSLINE::ComputeBoundingBox() {
|
|
int minx = INT32_MAX;
|
|
int miny = INT32_MAX;
|
|
int maxx = -INT32_MAX;
|
|
int maxy = -INT32_MAX;
|
|
|
|
// Find boundaries.
|
|
start = loop->pos;
|
|
EDGEPT* this_edge = loop;
|
|
do {
|
|
if (!this_edge->IsHidden() || !this_edge->prev->IsHidden()) {
|
|
if (this_edge->pos.x < minx)
|
|
minx = this_edge->pos.x;
|
|
if (this_edge->pos.y < miny)
|
|
miny = this_edge->pos.y;
|
|
if (this_edge->pos.x > maxx)
|
|
maxx = this_edge->pos.x;
|
|
if (this_edge->pos.y > maxy)
|
|
maxy = this_edge->pos.y;
|
|
}
|
|
this_edge = this_edge->next;
|
|
} while (this_edge != loop);
|
|
// Reset bounds.
|
|
topleft.x = minx;
|
|
topleft.y = maxy;
|
|
botright.x = maxx;
|
|
botright.y = miny;
|
|
}
|
|
|
|
// Computes the min and max cross product of the outline points with the
|
|
// given vec and returns the results in min_xp and max_xp. Geometrically
|
|
// this is the left and right edge of the outline perpendicular to the
|
|
// given direction, but to get the distance units correct, you would
|
|
// have to divide by the modulus of vec.
|
|
void TESSLINE::MinMaxCrossProduct(const TPOINT vec,
|
|
int* min_xp, int* max_xp) const {
|
|
*min_xp = INT32_MAX;
|
|
*max_xp = INT32_MIN;
|
|
EDGEPT* this_edge = loop;
|
|
do {
|
|
if (!this_edge->IsHidden() || !this_edge->prev->IsHidden()) {
|
|
int product = CROSS(this_edge->pos, vec);
|
|
UpdateRange(product, min_xp, max_xp);
|
|
}
|
|
this_edge = this_edge->next;
|
|
} while (this_edge != loop);
|
|
}
|
|
|
|
TBOX TESSLINE::bounding_box() const {
|
|
return TBOX(topleft.x, botright.y, botright.x, topleft.y);
|
|
}
|
|
|
|
#ifndef GRAPHICS_DISABLED
|
|
void TESSLINE::plot(ScrollView* window, ScrollView::Color color,
|
|
ScrollView::Color child_color) {
|
|
if (is_hole)
|
|
window->Pen(child_color);
|
|
else
|
|
window->Pen(color);
|
|
window->SetCursor(start.x, start.y);
|
|
EDGEPT* pt = loop;
|
|
do {
|
|
bool prev_hidden = pt->IsHidden();
|
|
pt = pt->next;
|
|
if (prev_hidden)
|
|
window->SetCursor(pt->pos.x, pt->pos.y);
|
|
else
|
|
window->DrawTo(pt->pos.x, pt->pos.y);
|
|
} while (pt != loop);
|
|
}
|
|
#endif // GRAPHICS_DISABLED
|
|
|
|
// Returns the first non-hidden EDGEPT that has a different src_outline to
|
|
// its predecessor, or, if all the same, the lowest indexed point.
|
|
EDGEPT* TESSLINE::FindBestStartPt() const {
|
|
EDGEPT* best_start = loop;
|
|
int best_step = loop->start_step;
|
|
// Iterate the polygon.
|
|
EDGEPT* pt = loop;
|
|
do {
|
|
if (pt->IsHidden()) continue;
|
|
if (pt->prev->IsHidden() || pt->prev->src_outline != pt->src_outline)
|
|
return pt; // Qualifies as the best.
|
|
if (pt->start_step < best_step) {
|
|
best_step = pt->start_step;
|
|
best_start = pt;
|
|
}
|
|
} while ((pt = pt->next) != loop);
|
|
return best_start;
|
|
}
|
|
|
|
// Iterate the given list of outlines, converting to TESSLINE by polygonal
|
|
// approximation and recursively any children, returning the current tail
|
|
// of the resulting list of TESSLINEs.
|
|
static TESSLINE** ApproximateOutlineList(bool allow_detailed_fx,
|
|
C_OUTLINE_LIST* outlines,
|
|
bool children,
|
|
TESSLINE** tail) {
|
|
C_OUTLINE_IT ol_it(outlines);
|
|
for (ol_it.mark_cycle_pt(); !ol_it.cycled_list(); ol_it.forward()) {
|
|
C_OUTLINE* outline = ol_it.data();
|
|
if (outline->pathlength() > 0) {
|
|
TESSLINE* tessline = ApproximateOutline(allow_detailed_fx, outline);
|
|
tessline->is_hole = children;
|
|
*tail = tessline;
|
|
tail = &tessline->next;
|
|
}
|
|
if (!outline->child()->empty()) {
|
|
tail = ApproximateOutlineList(allow_detailed_fx, outline->child(), true,
|
|
tail);
|
|
}
|
|
}
|
|
return tail;
|
|
}
|
|
|
|
// Factory to build a TBLOB from a C_BLOB with polygonal approximation along
|
|
// the way. If allow_detailed_fx is true, the EDGEPTs in the returned TBLOB
|
|
// contain pointers to the input C_OUTLINEs that enable higher-resolution
|
|
// feature extraction that does not use the polygonal approximation.
|
|
TBLOB* TBLOB::PolygonalCopy(bool allow_detailed_fx, C_BLOB* src) {
|
|
TBLOB* tblob = new TBLOB;
|
|
ApproximateOutlineList(allow_detailed_fx, src->out_list(), false,
|
|
&tblob->outlines);
|
|
return tblob;
|
|
}
|
|
|
|
// Factory builds a blob with no outlines, but copies the other member data.
|
|
TBLOB* TBLOB::ShallowCopy(const TBLOB& src) {
|
|
TBLOB* blob = new TBLOB;
|
|
blob->denorm_ = src.denorm_;
|
|
return blob;
|
|
}
|
|
|
|
// Normalizes the blob for classification only if needed.
|
|
// (Normally this means a non-zero classify rotation.)
|
|
// If no Normalization is needed, then nullptr is returned, and the input blob
|
|
// can be used directly. Otherwise a new TBLOB is returned which must be
|
|
// deleted after use.
|
|
TBLOB* TBLOB::ClassifyNormalizeIfNeeded() const {
|
|
TBLOB* rotated_blob = nullptr;
|
|
// If necessary, copy the blob and rotate it. The rotation is always
|
|
// +/- 90 degrees, as 180 was already taken care of.
|
|
if (denorm_.block() != nullptr &&
|
|
denorm_.block()->classify_rotation().y() != 0.0) {
|
|
TBOX box = bounding_box();
|
|
int x_middle = (box.left() + box.right()) / 2;
|
|
int y_middle = (box.top() + box.bottom()) / 2;
|
|
rotated_blob = new TBLOB(*this);
|
|
const FCOORD& rotation = denorm_.block()->classify_rotation();
|
|
// Move the rotated blob back to the same y-position so that we
|
|
// can still distinguish similar glyphs with differeny y-position.
|
|
float target_y = kBlnBaselineOffset +
|
|
(rotation.y() > 0 ? x_middle - box.left() : box.right() - x_middle);
|
|
rotated_blob->Normalize(nullptr, &rotation, &denorm_, x_middle, y_middle,
|
|
1.0f, 1.0f, 0.0f, target_y,
|
|
denorm_.inverse(), denorm_.pix());
|
|
}
|
|
return rotated_blob;
|
|
}
|
|
|
|
// Copies the data and the outline, but leaves next untouched.
|
|
void TBLOB::CopyFrom(const TBLOB& src) {
|
|
Clear();
|
|
TESSLINE* prev_outline = nullptr;
|
|
for (TESSLINE* srcline = src.outlines; srcline != nullptr;
|
|
srcline = srcline->next) {
|
|
TESSLINE* new_outline = new TESSLINE(*srcline);
|
|
if (outlines == nullptr)
|
|
outlines = new_outline;
|
|
else
|
|
prev_outline->next = new_outline;
|
|
prev_outline = new_outline;
|
|
}
|
|
denorm_ = src.denorm_;
|
|
}
|
|
|
|
// Deletes owned data.
|
|
void TBLOB::Clear() {
|
|
for (TESSLINE* next_outline = nullptr; outlines != nullptr;
|
|
outlines = next_outline) {
|
|
next_outline = outlines->next;
|
|
delete outlines;
|
|
}
|
|
}
|
|
|
|
// Sets up the built-in DENORM and normalizes the blob in-place.
|
|
// For parameters see DENORM::SetupNormalization, plus the inverse flag for
|
|
// this blob and the Pix for the full image.
|
|
void TBLOB::Normalize(const BLOCK* block,
|
|
const FCOORD* rotation,
|
|
const DENORM* predecessor,
|
|
float x_origin, float y_origin,
|
|
float x_scale, float y_scale,
|
|
float final_xshift, float final_yshift,
|
|
bool inverse, Pix* pix) {
|
|
denorm_.SetupNormalization(block, rotation, predecessor, x_origin, y_origin,
|
|
x_scale, y_scale, final_xshift, final_yshift);
|
|
denorm_.set_inverse(inverse);
|
|
denorm_.set_pix(pix);
|
|
// TODO(rays) outline->Normalize is more accurate, but breaks tests due
|
|
// the changes it makes. Reinstate this code with a retraining.
|
|
// The reason this change is troublesome is that it normalizes for the
|
|
// baseline value computed independently at each x-coord. If the baseline
|
|
// is not horizontal, this introduces shear into the normalized blob, which
|
|
// is useful on the rare occasions that the baseline is really curved, but
|
|
// the baselines need to be stabilized the rest of the time.
|
|
#if 0
|
|
for (TESSLINE* outline = outlines; outline != nullptr; outline = outline->next) {
|
|
outline->Normalize(denorm_);
|
|
}
|
|
#else
|
|
denorm_.LocalNormBlob(this);
|
|
#endif
|
|
}
|
|
|
|
// Rotates by the given rotation in place.
|
|
void TBLOB::Rotate(const FCOORD rotation) {
|
|
for (TESSLINE* outline = outlines; outline != nullptr; outline = outline->next) {
|
|
outline->Rotate(rotation);
|
|
}
|
|
}
|
|
|
|
// Moves by the given vec in place.
|
|
void TBLOB::Move(const ICOORD vec) {
|
|
for (TESSLINE* outline = outlines; outline != nullptr; outline = outline->next) {
|
|
outline->Move(vec);
|
|
}
|
|
}
|
|
|
|
// Scales by the given factor in place.
|
|
void TBLOB::Scale(float factor) {
|
|
for (TESSLINE* outline = outlines; outline != nullptr; outline = outline->next) {
|
|
outline->Scale(factor);
|
|
}
|
|
}
|
|
|
|
// Recomputes the bounding boxes of the outlines.
|
|
void TBLOB::ComputeBoundingBoxes() {
|
|
for (TESSLINE* outline = outlines; outline != nullptr; outline = outline->next) {
|
|
outline->ComputeBoundingBox();
|
|
}
|
|
}
|
|
|
|
// Returns the number of outlines.
|
|
int TBLOB::NumOutlines() const {
|
|
int result = 0;
|
|
for (TESSLINE* outline = outlines; outline != nullptr; outline = outline->next)
|
|
++result;
|
|
return result;
|
|
}
|
|
|
|
/**********************************************************************
|
|
* TBLOB::bounding_box()
|
|
*
|
|
* Compute the bounding_box of a compound blob, defined to be the
|
|
* bounding box of the union of all top-level outlines in the blob.
|
|
**********************************************************************/
|
|
TBOX TBLOB::bounding_box() const {
|
|
if (outlines == nullptr)
|
|
return TBOX(0, 0, 0, 0);
|
|
TESSLINE *outline = outlines;
|
|
TBOX box = outline->bounding_box();
|
|
for (outline = outline->next; outline != nullptr; outline = outline->next) {
|
|
box += outline->bounding_box();
|
|
}
|
|
return box;
|
|
}
|
|
|
|
// Finds and deletes any duplicate outlines in this blob, without deleting
|
|
// their EDGEPTs.
|
|
void TBLOB::EliminateDuplicateOutlines() {
|
|
for (TESSLINE* outline = outlines; outline != nullptr; outline = outline->next) {
|
|
TESSLINE* last_outline = outline;
|
|
for (TESSLINE* other_outline = outline->next; other_outline != nullptr;
|
|
last_outline = other_outline, other_outline = other_outline->next) {
|
|
if (outline->SameBox(*other_outline)) {
|
|
last_outline->next = other_outline->next;
|
|
// This doesn't leak - the outlines share the EDGEPTs.
|
|
other_outline->loop = nullptr;
|
|
delete other_outline;
|
|
other_outline = last_outline;
|
|
// If it is part of a cut, then it can't be a hole any more.
|
|
outline->is_hole = false;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// Swaps the outlines of *this and next if needed to keep the centers in
|
|
// increasing x.
|
|
void TBLOB::CorrectBlobOrder(TBLOB* next) {
|
|
TBOX box = bounding_box();
|
|
TBOX next_box = next->bounding_box();
|
|
if (box.x_middle() > next_box.x_middle()) {
|
|
Swap(&outlines, &next->outlines);
|
|
}
|
|
}
|
|
|
|
#ifndef GRAPHICS_DISABLED
|
|
void TBLOB::plot(ScrollView* window, ScrollView::Color color,
|
|
ScrollView::Color child_color) {
|
|
for (TESSLINE* outline = outlines; outline != nullptr; outline = outline->next)
|
|
outline->plot(window, color, child_color);
|
|
}
|
|
#endif // GRAPHICS_DISABLED
|
|
|
|
// Computes the center of mass and second moments for the old baseline and
|
|
// 2nd moment normalizations. Returns the outline length.
|
|
// The input denorm should be the normalizations that have been applied from
|
|
// the image to the current state of this TBLOB.
|
|
int TBLOB::ComputeMoments(FCOORD* center, FCOORD* second_moments) const {
|
|
// Compute 1st and 2nd moments of the original outline.
|
|
LLSQ accumulator;
|
|
TBOX box = bounding_box();
|
|
// Iterate the outlines, accumulating edges relative the box.botleft().
|
|
CollectEdges(box, nullptr, &accumulator, nullptr, nullptr);
|
|
*center = accumulator.mean_point() + box.botleft();
|
|
// The 2nd moments are just the standard deviation of the point positions.
|
|
double x2nd = sqrt(accumulator.x_variance());
|
|
double y2nd = sqrt(accumulator.y_variance());
|
|
if (x2nd < 1.0) x2nd = 1.0;
|
|
if (y2nd < 1.0) y2nd = 1.0;
|
|
second_moments->set_x(x2nd);
|
|
second_moments->set_y(y2nd);
|
|
return accumulator.count();
|
|
}
|
|
|
|
// Computes the precise bounding box of the coords that are generated by
|
|
// GetEdgeCoords. This may be different from the bounding box of the polygon.
|
|
void TBLOB::GetPreciseBoundingBox(TBOX* precise_box) const {
|
|
TBOX box = bounding_box();
|
|
*precise_box = TBOX();
|
|
CollectEdges(box, precise_box, nullptr, nullptr, nullptr);
|
|
precise_box->move(box.botleft());
|
|
}
|
|
|
|
// Adds edges to the given vectors.
|
|
// For all the edge steps in all the outlines, or polygonal approximation
|
|
// where there are no edge steps, collects the steps into x_coords/y_coords.
|
|
// x_coords is a collection of the x-coords of vertical edges for each
|
|
// y-coord starting at box.bottom().
|
|
// y_coords is a collection of the y-coords of horizontal edges for each
|
|
// x-coord starting at box.left().
|
|
// Eg x_coords[0] is a collection of the x-coords of edges at y=bottom.
|
|
// Eg x_coords[1] is a collection of the x-coords of edges at y=bottom + 1.
|
|
void TBLOB::GetEdgeCoords(const TBOX& box,
|
|
GenericVector<GenericVector<int> >* x_coords,
|
|
GenericVector<GenericVector<int> >* y_coords) const {
|
|
GenericVector<int> empty;
|
|
x_coords->init_to_size(box.height(), empty);
|
|
y_coords->init_to_size(box.width(), empty);
|
|
CollectEdges(box, nullptr, nullptr, x_coords, y_coords);
|
|
// Sort the output vectors.
|
|
for (int i = 0; i < x_coords->size(); ++i)
|
|
(*x_coords)[i].sort();
|
|
for (int i = 0; i < y_coords->size(); ++i)
|
|
(*y_coords)[i].sort();
|
|
}
|
|
|
|
// Accumulates the segment between pt1 and pt2 in the LLSQ, quantizing over
|
|
// the integer coordinate grid to properly weight long vectors.
|
|
static void SegmentLLSQ(const FCOORD& pt1, const FCOORD& pt2,
|
|
LLSQ* accumulator) {
|
|
FCOORD step(pt2);
|
|
step -= pt1;
|
|
int xstart = IntCastRounded(MIN(pt1.x(), pt2.x()));
|
|
int xend = IntCastRounded(MAX(pt1.x(), pt2.x()));
|
|
int ystart = IntCastRounded(MIN(pt1.y(), pt2.y()));
|
|
int yend = IntCastRounded(MAX(pt1.y(), pt2.y()));
|
|
if (xstart == xend && ystart == yend) return; // Nothing to do.
|
|
double weight = step.length() / (xend - xstart + yend - ystart);
|
|
// Compute and save the y-position at the middle of each x-step.
|
|
for (int x = xstart; x < xend; ++x) {
|
|
double y = pt1.y() + step.y() * (x + 0.5 - pt1.x()) / step.x();
|
|
accumulator->add(x + 0.5, y, weight);
|
|
}
|
|
// Compute and save the x-position at the middle of each y-step.
|
|
for (int y = ystart; y < yend; ++y) {
|
|
double x = pt1.x() + step.x() * (y + 0.5 - pt1.y()) / step.y();
|
|
accumulator->add(x, y + 0.5, weight);
|
|
}
|
|
}
|
|
|
|
// Adds any edges from a single segment of outline between pt1 and pt2 to
|
|
// the x_coords, y_coords vectors. pt1 and pt2 should be relative to the
|
|
// bottom-left of the bounding box, hence indices to x_coords, y_coords
|
|
// are clipped to ([0,x_limit], [0,y_limit]).
|
|
// See GetEdgeCoords above for a description of x_coords, y_coords.
|
|
static void SegmentCoords(const FCOORD& pt1, const FCOORD& pt2,
|
|
int x_limit, int y_limit,
|
|
GenericVector<GenericVector<int> >* x_coords,
|
|
GenericVector<GenericVector<int> >* y_coords) {
|
|
FCOORD step(pt2);
|
|
step -= pt1;
|
|
int start = ClipToRange(IntCastRounded(MIN(pt1.x(), pt2.x())), 0, x_limit);
|
|
int end = ClipToRange(IntCastRounded(MAX(pt1.x(), pt2.x())), 0, x_limit);
|
|
for (int x = start; x < end; ++x) {
|
|
int y = IntCastRounded(pt1.y() + step.y() * (x + 0.5 - pt1.x()) / step.x());
|
|
(*y_coords)[x].push_back(y);
|
|
}
|
|
start = ClipToRange(IntCastRounded(MIN(pt1.y(), pt2.y())), 0, y_limit);
|
|
end = ClipToRange(IntCastRounded(MAX(pt1.y(), pt2.y())), 0, y_limit);
|
|
for (int y = start; y < end; ++y) {
|
|
int x = IntCastRounded(pt1.x() + step.x() * (y + 0.5 - pt1.y()) / step.y());
|
|
(*x_coords)[y].push_back(x);
|
|
}
|
|
}
|
|
|
|
// Adds any edges from a single segment of outline between pt1 and pt2 to
|
|
// the bbox such that it guarantees to contain anything produced by
|
|
// SegmentCoords.
|
|
static void SegmentBBox(const FCOORD& pt1, const FCOORD& pt2, TBOX* bbox) {
|
|
FCOORD step(pt2);
|
|
step -= pt1;
|
|
int x1 = IntCastRounded(MIN(pt1.x(), pt2.x()));
|
|
int x2 = IntCastRounded(MAX(pt1.x(), pt2.x()));
|
|
if (x2 > x1) {
|
|
int y1 = IntCastRounded(pt1.y() + step.y() * (x1 + 0.5 - pt1.x()) /
|
|
step.x());
|
|
int y2 = IntCastRounded(pt1.y() + step.y() * (x2 - 0.5 - pt1.x()) /
|
|
step.x());
|
|
TBOX point(x1, MIN(y1, y2), x2, MAX(y1, y2));
|
|
*bbox += point;
|
|
}
|
|
int y1 = IntCastRounded(MIN(pt1.y(), pt2.y()));
|
|
int y2 = IntCastRounded(MAX(pt1.y(), pt2.y()));
|
|
if (y2 > y1) {
|
|
int x1 = IntCastRounded(pt1.x() + step.x() * (y1 + 0.5 - pt1.y()) /
|
|
step.y());
|
|
int x2 = IntCastRounded(pt1.x() + step.x() * (y2 - 0.5 - pt1.y()) /
|
|
step.y());
|
|
TBOX point(MIN(x1, x2), y1, MAX(x1, x2), y2);
|
|
*bbox += point;
|
|
}
|
|
}
|
|
|
|
// Collects edges into the given bounding box, LLSQ accumulator and/or x_coords,
|
|
// y_coords vectors.
|
|
// For a description of x_coords/y_coords, see GetEdgeCoords above.
|
|
// Startpt to lastpt, inclusive, MUST have the same src_outline member,
|
|
// which may be nullptr. The vector from lastpt to its next is included in
|
|
// the accumulation. Hidden edges should be excluded by the caller.
|
|
// The input denorm should be the normalizations that have been applied from
|
|
// the image to the current state of the TBLOB from which startpt, lastpt come.
|
|
// box is the bounding box of the blob from which the EDGEPTs are taken and
|
|
// indices into x_coords, y_coords are offset by box.botleft().
|
|
static void CollectEdgesOfRun(const EDGEPT* startpt, const EDGEPT* lastpt,
|
|
const DENORM& denorm, const TBOX& box,
|
|
TBOX* bounding_box,
|
|
LLSQ* accumulator,
|
|
GenericVector<GenericVector<int> > *x_coords,
|
|
GenericVector<GenericVector<int> > *y_coords) {
|
|
const C_OUTLINE* outline = startpt->src_outline;
|
|
int x_limit = box.width() - 1;
|
|
int y_limit = box.height() - 1;
|
|
if (outline != nullptr) {
|
|
// Use higher-resolution edge points stored on the outline.
|
|
// The outline coordinates may not match the binary image because of the
|
|
// rotation for vertical text lines, but the root_denorm IS the matching
|
|
// start of the DENORM chain.
|
|
const DENORM* root_denorm = denorm.RootDenorm();
|
|
int step_length = outline->pathlength();
|
|
int start_index = startpt->start_step;
|
|
// Note that if this run straddles the wrap-around point of the outline,
|
|
// that lastpt->start_step may have a lower index than startpt->start_step,
|
|
// and we want to use an end_index that allows us to use a positive
|
|
// increment, so we add step_length if necessary, but that may be beyond the
|
|
// bounds of the outline steps/ due to wrap-around, so we use % step_length
|
|
// everywhere, except for start_index.
|
|
int end_index = lastpt->start_step + lastpt->step_count;
|
|
if (end_index <= start_index)
|
|
end_index += step_length;
|
|
// pos is the integer coordinates of the binary image steps.
|
|
ICOORD pos = outline->position_at_index(start_index);
|
|
FCOORD origin(box.left(), box.bottom());
|
|
// f_pos is a floating-point version of pos that offers improved edge
|
|
// positioning using greyscale information or smoothing of edge steps.
|
|
FCOORD f_pos = outline->sub_pixel_pos_at_index(pos, start_index);
|
|
// pos_normed is f_pos after the appropriate normalization, and relative
|
|
// to origin.
|
|
// prev_normed is the previous value of pos_normed.
|
|
FCOORD prev_normed;
|
|
denorm.NormTransform(root_denorm, f_pos, &prev_normed);
|
|
prev_normed -= origin;
|
|
for (int index = start_index; index < end_index; ++index) {
|
|
ICOORD step = outline->step(index % step_length);
|
|
// Only use the point if its edge strength is positive. This excludes
|
|
// points that don't provide useful information, eg
|
|
// ___________
|
|
// |___________
|
|
// The vertical step provides only noisy, damaging information, as even
|
|
// with a greyscale image, the positioning of the edge there may be a
|
|
// fictitious extrapolation, so previous processing has eliminated it.
|
|
if (outline->edge_strength_at_index(index % step_length) > 0) {
|
|
FCOORD f_pos = outline->sub_pixel_pos_at_index(pos,
|
|
index % step_length);
|
|
FCOORD pos_normed;
|
|
denorm.NormTransform(root_denorm, f_pos, &pos_normed);
|
|
pos_normed -= origin;
|
|
// Accumulate the information that is selected by the caller.
|
|
if (bounding_box != nullptr) {
|
|
SegmentBBox(pos_normed, prev_normed, bounding_box);
|
|
}
|
|
if (accumulator != nullptr) {
|
|
SegmentLLSQ(pos_normed, prev_normed, accumulator);
|
|
}
|
|
if (x_coords != nullptr && y_coords != nullptr) {
|
|
SegmentCoords(pos_normed, prev_normed, x_limit, y_limit,
|
|
x_coords, y_coords);
|
|
}
|
|
prev_normed = pos_normed;
|
|
}
|
|
pos += step;
|
|
}
|
|
} else {
|
|
// There is no outline, so we are forced to use the polygonal approximation.
|
|
const EDGEPT* endpt = lastpt->next;
|
|
const EDGEPT* pt = startpt;
|
|
do {
|
|
FCOORD next_pos(pt->next->pos.x - box.left(),
|
|
pt->next->pos.y - box.bottom());
|
|
FCOORD pos(pt->pos.x - box.left(), pt->pos.y - box.bottom());
|
|
if (bounding_box != nullptr) {
|
|
SegmentBBox(next_pos, pos, bounding_box);
|
|
}
|
|
if (accumulator != nullptr) {
|
|
SegmentLLSQ(next_pos, pos, accumulator);
|
|
}
|
|
if (x_coords != nullptr && y_coords != nullptr) {
|
|
SegmentCoords(next_pos, pos, x_limit, y_limit, x_coords, y_coords);
|
|
}
|
|
} while ((pt = pt->next) != endpt);
|
|
}
|
|
}
|
|
|
|
// For all the edge steps in all the outlines, or polygonal approximation
|
|
// where there are no edge steps, collects the steps into the bounding_box,
|
|
// llsq and/or the x_coords/y_coords. Both are used in different kinds of
|
|
// normalization.
|
|
// For a description of x_coords, y_coords, see GetEdgeCoords above.
|
|
void TBLOB::CollectEdges(const TBOX& box,
|
|
TBOX* bounding_box, LLSQ* llsq,
|
|
GenericVector<GenericVector<int> >* x_coords,
|
|
GenericVector<GenericVector<int> >* y_coords) const {
|
|
// Iterate the outlines.
|
|
for (const TESSLINE* ol = outlines; ol != nullptr; ol = ol->next) {
|
|
// Iterate the polygon.
|
|
EDGEPT* loop_pt = ol->FindBestStartPt();
|
|
EDGEPT* pt = loop_pt;
|
|
if (pt == nullptr) continue;
|
|
do {
|
|
if (pt->IsHidden()) continue;
|
|
// Find a run of equal src_outline.
|
|
EDGEPT* last_pt = pt;
|
|
do {
|
|
last_pt = last_pt->next;
|
|
} while (last_pt != loop_pt && !last_pt->IsHidden() &&
|
|
last_pt->src_outline == pt->src_outline);
|
|
last_pt = last_pt->prev;
|
|
CollectEdgesOfRun(pt, last_pt, denorm_, box,
|
|
bounding_box, llsq, x_coords, y_coords);
|
|
pt = last_pt;
|
|
} while ((pt = pt->next) != loop_pt);
|
|
}
|
|
}
|
|
|
|
// Factory to build a TWERD from a (C_BLOB) WERD, with polygonal
|
|
// approximation along the way.
|
|
TWERD* TWERD::PolygonalCopy(bool allow_detailed_fx, WERD* src) {
|
|
TWERD* tessword = new TWERD;
|
|
tessword->latin_script = src->flag(W_SCRIPT_IS_LATIN);
|
|
C_BLOB_IT b_it(src->cblob_list());
|
|
for (b_it.mark_cycle_pt(); !b_it.cycled_list(); b_it.forward()) {
|
|
C_BLOB* blob = b_it.data();
|
|
TBLOB* tblob = TBLOB::PolygonalCopy(allow_detailed_fx, blob);
|
|
tessword->blobs.push_back(tblob);
|
|
}
|
|
return tessword;
|
|
}
|
|
|
|
// Baseline normalizes the blobs in-place, recording the normalization in the
|
|
// DENORMs in the blobs.
|
|
void TWERD::BLNormalize(const BLOCK* block, const ROW* row, Pix* pix,
|
|
bool inverse, float x_height, float baseline_shift,
|
|
bool numeric_mode, tesseract::OcrEngineMode hint,
|
|
const TBOX* norm_box,
|
|
DENORM* word_denorm) {
|
|
TBOX word_box = bounding_box();
|
|
if (norm_box != nullptr) word_box = *norm_box;
|
|
float word_middle = (word_box.left() + word_box.right()) / 2.0f;
|
|
float input_y_offset = 0.0f;
|
|
float final_y_offset = static_cast<float>(kBlnBaselineOffset);
|
|
float scale = kBlnXHeight / x_height;
|
|
if (row == nullptr) {
|
|
word_middle = word_box.left();
|
|
input_y_offset = word_box.bottom();
|
|
final_y_offset = 0.0f;
|
|
} else {
|
|
input_y_offset = row->base_line(word_middle) + baseline_shift;
|
|
}
|
|
for (int b = 0; b < blobs.size(); ++b) {
|
|
TBLOB* blob = blobs[b];
|
|
TBOX blob_box = blob->bounding_box();
|
|
float mid_x = (blob_box.left() + blob_box.right()) / 2.0f;
|
|
float baseline = input_y_offset;
|
|
float blob_scale = scale;
|
|
if (numeric_mode) {
|
|
baseline = blob_box.bottom();
|
|
blob_scale = ClipToRange(kBlnXHeight * 4.0f / (3 * blob_box.height()),
|
|
scale, scale * 1.5f);
|
|
} else if (row != nullptr) {
|
|
baseline = row->base_line(mid_x) + baseline_shift;
|
|
}
|
|
// The image will be 8-bit grey if the input was grey or color. Note that in
|
|
// a grey image 0 is black and 255 is white. If the input was binary, then
|
|
// the pix will be binary and 0 is white, with 1 being black.
|
|
// To tell the difference pixGetDepth() will return 8 or 1.
|
|
// The inverse flag will be true iff the word has been determined to be
|
|
// white on black, and is independent of whether the pix is 8 bit or 1 bit.
|
|
blob->Normalize(block, nullptr, nullptr, word_middle, baseline, blob_scale,
|
|
blob_scale, 0.0f, final_y_offset, inverse, pix);
|
|
}
|
|
if (word_denorm != nullptr) {
|
|
word_denorm->SetupNormalization(block, nullptr, nullptr, word_middle,
|
|
input_y_offset, scale, scale,
|
|
0.0f, final_y_offset);
|
|
word_denorm->set_inverse(inverse);
|
|
word_denorm->set_pix(pix);
|
|
}
|
|
}
|
|
|
|
// Copies the data and the blobs, but leaves next untouched.
|
|
void TWERD::CopyFrom(const TWERD& src) {
|
|
Clear();
|
|
latin_script = src.latin_script;
|
|
for (int b = 0; b < src.blobs.size(); ++b) {
|
|
TBLOB* new_blob = new TBLOB(*src.blobs[b]);
|
|
blobs.push_back(new_blob);
|
|
}
|
|
}
|
|
|
|
// Deletes owned data.
|
|
void TWERD::Clear() {
|
|
blobs.delete_data_pointers();
|
|
blobs.clear();
|
|
}
|
|
|
|
// Recomputes the bounding boxes of the blobs.
|
|
void TWERD::ComputeBoundingBoxes() {
|
|
for (int b = 0; b < blobs.size(); ++b) {
|
|
blobs[b]->ComputeBoundingBoxes();
|
|
}
|
|
}
|
|
|
|
TBOX TWERD::bounding_box() const {
|
|
TBOX result;
|
|
for (int b = 0; b < blobs.size(); ++b) {
|
|
TBOX box = blobs[b]->bounding_box();
|
|
result += box;
|
|
}
|
|
return result;
|
|
}
|
|
|
|
// Merges the blobs from start to end, not including end, and deletes
|
|
// the blobs between start and end.
|
|
void TWERD::MergeBlobs(int start, int end) {
|
|
if (start >= blobs.size() - 1) return; // Nothing to do.
|
|
TESSLINE* outline = blobs[start]->outlines;
|
|
for (int i = start + 1; i < end && i < blobs.size(); ++i) {
|
|
TBLOB* next_blob = blobs[i];
|
|
// Take the outlines from the next blob.
|
|
if (outline == nullptr) {
|
|
blobs[start]->outlines = next_blob->outlines;
|
|
outline = blobs[start]->outlines;
|
|
} else {
|
|
while (outline->next != nullptr)
|
|
outline = outline->next;
|
|
outline->next = next_blob->outlines;
|
|
next_blob->outlines = nullptr;
|
|
}
|
|
// Delete the next blob and move on.
|
|
delete next_blob;
|
|
blobs[i] = nullptr;
|
|
}
|
|
// Remove dead blobs from the vector.
|
|
for (int i = start + 1; i < end && start + 1 < blobs.size(); ++i) {
|
|
blobs.remove(start + 1);
|
|
}
|
|
}
|
|
|
|
#ifndef GRAPHICS_DISABLED
|
|
void TWERD::plot(ScrollView* window) {
|
|
ScrollView::Color color = WERD::NextColor(ScrollView::BLACK);
|
|
for (int b = 0; b < blobs.size(); ++b) {
|
|
blobs[b]->plot(window, color, ScrollView::BROWN);
|
|
color = WERD::NextColor(color);
|
|
}
|
|
}
|
|
#endif // GRAPHICS_DISABLED
|
|
|
|
/**********************************************************************
|
|
* divisible_blob
|
|
*
|
|
* Returns true if the blob contains multiple outlines than can be
|
|
* separated using divide_blobs. Sets the location to be used in the
|
|
* call to divide_blobs.
|
|
**********************************************************************/
|
|
bool divisible_blob(TBLOB *blob, bool italic_blob, TPOINT* location) {
|
|
if (blob->outlines == nullptr || blob->outlines->next == nullptr)
|
|
return false; // Need at least 2 outlines for it to be possible.
|
|
int max_gap = 0;
|
|
TPOINT vertical = italic_blob ? kDivisibleVerticalItalic
|
|
: kDivisibleVerticalUpright;
|
|
for (TESSLINE* outline1 = blob->outlines; outline1 != nullptr;
|
|
outline1 = outline1->next) {
|
|
if (outline1->is_hole)
|
|
continue; // Holes do not count as separable.
|
|
TPOINT mid_pt1(
|
|
static_cast<int16_t>((outline1->topleft.x + outline1->botright.x) / 2),
|
|
static_cast<int16_t>((outline1->topleft.y + outline1->botright.y) / 2));
|
|
int mid_prod1 = CROSS(mid_pt1, vertical);
|
|
int min_prod1, max_prod1;
|
|
outline1->MinMaxCrossProduct(vertical, &min_prod1, &max_prod1);
|
|
for (TESSLINE* outline2 = outline1->next; outline2 != nullptr;
|
|
outline2 = outline2->next) {
|
|
if (outline2->is_hole)
|
|
continue; // Holes do not count as separable.
|
|
TPOINT mid_pt2(
|
|
static_cast<int16_t>((outline2->topleft.x + outline2->botright.x) / 2),
|
|
static_cast<int16_t>((outline2->topleft.y + outline2->botright.y) / 2));
|
|
int mid_prod2 = CROSS(mid_pt2, vertical);
|
|
int min_prod2, max_prod2;
|
|
outline2->MinMaxCrossProduct(vertical, &min_prod2, &max_prod2);
|
|
int mid_gap = abs(mid_prod2 - mid_prod1);
|
|
int overlap = MIN(max_prod1, max_prod2) - MAX(min_prod1, min_prod2);
|
|
if (mid_gap - overlap / 4 > max_gap) {
|
|
max_gap = mid_gap - overlap / 4;
|
|
*location = mid_pt1;
|
|
*location += mid_pt2;
|
|
*location /= 2;
|
|
}
|
|
}
|
|
}
|
|
// Use the y component of the vertical vector as an approximation to its
|
|
// length.
|
|
return max_gap > vertical.y;
|
|
}
|
|
|
|
/**********************************************************************
|
|
* divide_blobs
|
|
*
|
|
* Create two blobs by grouping the outlines in the appropriate blob.
|
|
* The outlines that are beyond the location point are moved to the
|
|
* other blob. The ones whose x location is less than that point are
|
|
* retained in the original blob.
|
|
**********************************************************************/
|
|
void divide_blobs(TBLOB *blob, TBLOB *other_blob, bool italic_blob,
|
|
const TPOINT& location) {
|
|
TPOINT vertical = italic_blob ? kDivisibleVerticalItalic
|
|
: kDivisibleVerticalUpright;
|
|
TESSLINE *outline1 = nullptr;
|
|
TESSLINE *outline2 = nullptr;
|
|
|
|
TESSLINE *outline = blob->outlines;
|
|
blob->outlines = nullptr;
|
|
int location_prod = CROSS(location, vertical);
|
|
|
|
while (outline != nullptr) {
|
|
TPOINT mid_pt(
|
|
static_cast<int16_t>((outline->topleft.x + outline->botright.x) / 2),
|
|
static_cast<int16_t>((outline->topleft.y + outline->botright.y) / 2));
|
|
int mid_prod = CROSS(mid_pt, vertical);
|
|
if (mid_prod < location_prod) {
|
|
// Outline is in left blob.
|
|
if (outline1)
|
|
outline1->next = outline;
|
|
else
|
|
blob->outlines = outline;
|
|
outline1 = outline;
|
|
} else {
|
|
// Outline is in right blob.
|
|
if (outline2)
|
|
outline2->next = outline;
|
|
else
|
|
other_blob->outlines = outline;
|
|
outline2 = outline;
|
|
}
|
|
outline = outline->next;
|
|
}
|
|
|
|
if (outline1)
|
|
outline1->next = nullptr;
|
|
if (outline2)
|
|
outline2->next = nullptr;
|
|
}
|