| 1 |
/*************************************************************************** |
| 2 |
* Copyright (c) 2009 Enrico Ros * |
| 3 |
* 2009 Enrico Ros <enrico.ros@email.it> * |
| 4 |
* 2009 Alberto Scarpa <skaal.sl@gmail.com> * |
| 5 |
* * |
| 6 |
* Permission is hereby granted, free of charge, to any person * |
| 7 |
* obtaining a copy of this software and associated documentation * |
| 8 |
* files (the "Software"), to deal in the Software without * |
| 9 |
* restriction, including without limitation the rights to use, * |
| 10 |
* copy, modify, merge, publish, distribute, sublicense, and/or sell * |
| 11 |
* copies of the Software, and to permit persons to whom the * |
| 12 |
* Software is furnished to do so, subject to the following * |
| 13 |
* conditions: * |
| 14 |
* * |
| 15 |
* The above copyright notice and this permission notice shall be * |
| 16 |
* included in all copies or substantial portions of the Software. * |
| 17 |
* * |
| 18 |
***************************************************************************/ |
| 19 |
|
| 20 |
#include "Ocr.h" |
| 21 |
#include <QFontMetrics> |
| 22 |
#include <QPainter> |
| 23 |
#include <QBitmap> |
| 24 |
#include <QTransform> |
| 25 |
#include <QPixmap> |
| 26 |
#include <math.h> |
| 27 |
|
| 28 |
#include <QDebug> |
| 29 |
#include <QLabel> |
| 30 |
|
| 31 |
struct OcrGlyph { |
| 32 |
QChar character; |
| 33 |
QImage image; |
| 34 |
int width; |
| 35 |
int height; |
| 36 |
double ratio; |
| 37 |
// ADD more and more properties ... |
| 38 |
}; |
| 39 |
|
| 40 |
|
| 41 |
// CORE FUNCTIONS |
| 42 |
static QImage trimImage( const QImage & image ) |
| 43 |
{ |
| 44 |
// find TOP |
| 45 |
int W = image.width(); |
| 46 |
int H = image.height(); |
| 47 |
int left = 0, top = 0, right = W - 1, bottom = H - 1; |
| 48 |
|
| 49 |
// find -left- margin |
| 50 |
bool blank = true; |
| 51 |
while ( blank && left < W ) { |
| 52 |
for ( int y = 0; y < H; y++ ) { |
| 53 |
if ( qGray( image.pixel( left, y ) ) < 32 ) { |
| 54 |
blank = false; |
| 55 |
break; |
| 56 |
} |
| 57 |
} |
| 58 |
if ( blank ) |
| 59 |
left++; |
| 60 |
} |
| 61 |
|
| 62 |
// find -top- margin |
| 63 |
blank = true; |
| 64 |
while ( blank && top < H ) { |
| 65 |
for ( int x = 0; x < W; x++ ) { |
| 66 |
if ( qGray( image.pixel( x, top ) ) < 32 ) { |
| 67 |
blank = false; |
| 68 |
break; |
| 69 |
} |
| 70 |
} |
| 71 |
if ( blank ) |
| 72 |
top++; |
| 73 |
} |
| 74 |
|
| 75 |
// find -right- margin |
| 76 |
blank = true; |
| 77 |
while ( blank && right >= 0 ) { |
| 78 |
for ( int y = 0; y < H; y++ ) { |
| 79 |
if ( qGray( image.pixel( right, y ) ) < 32 ) { |
| 80 |
blank = false; |
| 81 |
break; |
| 82 |
} |
| 83 |
} |
| 84 |
if ( blank ) |
| 85 |
right--; |
| 86 |
} |
| 87 |
|
| 88 |
// find -bottom- margin |
| 89 |
blank = true; |
| 90 |
while ( blank && bottom >= 0 ) { |
| 91 |
for ( int x = 0; x < W; x++ ) { |
| 92 |
if ( qGray( image.pixel( x, bottom ) ) < 32 ) { |
| 93 |
blank = false; |
| 94 |
break; |
| 95 |
} |
| 96 |
} |
| 97 |
if ( blank ) |
| 98 |
bottom--; |
| 99 |
} |
| 100 |
|
| 101 |
// ok: return cropped image |
| 102 |
W = right - left + 1; |
| 103 |
H = bottom - top + 1; |
| 104 |
if ( W > 0 && H > 0 ) |
| 105 |
return image.copy( left, top, W, H ); |
| 106 |
|
| 107 |
// invalid: return the same image |
| 108 |
qWarning( "trimImage: invalid image of size %dx%d", image.width(), image.height() ); |
| 109 |
return image; |
| 110 |
} |
| 111 |
|
| 112 |
static OcrResult compareGlyph( const QImage & image, OcrGlyph * glyph ) |
| 113 |
{ |
| 114 |
OcrResult result; |
| 115 |
result.character = glyph->character; |
| 116 |
result.confidence = 0.0; |
| 117 |
|
| 118 |
// bail out if ratios are too different |
| 119 |
double ratio = (float)image.width() / (float)image.height(); |
| 120 |
double k2 = (ratio / glyph->ratio + glyph->ratio / ratio) / 2; |
| 121 |
if ( k2 > 1.5 ) |
| 122 |
return result; |
| 123 |
|
| 124 |
// ALGO |
| 125 |
// scale glyph image size to image size |
| 126 |
#if 0 |
| 127 |
QTransform t; |
| 128 |
t.scale( (float)glyph->image.width() / (float)image.width(), (float)glyph->image.height() / (float)image.height() ); |
| 129 |
QImage glyphImage = glyph->image.transformed( t.inverted(0), Qt::SmoothTransformation ); |
| 130 |
#endif |
| 131 |
QImage glyphImage = glyph->image.scaled( image.width(), image.height() ); |
| 132 |
|
| 133 |
int width = glyphImage.width(); |
| 134 |
int height = glyphImage.height(); |
| 135 |
//int pixels = width * height; |
| 136 |
unsigned int error = 0; |
| 137 |
unsigned int analyzed = 0; |
| 138 |
for ( int y = 0; y < height; y++ ) { |
| 139 |
for ( int x = 0; x < width; x++ ) { |
| 140 |
int v1 = qGray( image.pixel( x, y ) ); |
| 141 |
int v2 = qGray( glyphImage.pixel( x, y ) ); |
| 142 |
if ( v1 >= 254 && v2 >= 254 ) |
| 143 |
continue; |
| 144 |
analyzed++; |
| 145 |
int diff = (v1 > v2 ? v1 - v2 : v2 - v1); |
| 146 |
error += (diff * diff) >> 8; |
| 147 |
} |
| 148 |
} |
| 149 |
if ( !analyzed ) |
| 150 |
return result; |
| 151 |
float KX = sqrt( (float)error / (254.0 * (float)analyzed) ); |
| 152 |
|
| 153 |
#if 0 |
| 154 |
if ( glyph->character == 'R' ) { |
| 155 |
QLabel * label = new QLabel(); |
| 156 |
label->setPixmap( QPixmap::fromImage( image ) ); |
| 157 |
label->show(); |
| 158 |
|
| 159 |
label = new QLabel(); |
| 160 |
label->setPixmap( QPixmap::fromImage( glyphImage ) ); |
| 161 |
label->show(); |
| 162 |
} |
| 163 |
#endif |
| 164 |
|
| 165 |
result.confidence = 1.0 - KX; |
| 166 |
//qWarning() << result.confidence << result.character << pixels << analyzed; |
| 167 |
return result; |
| 168 |
} |
| 169 |
|
| 170 |
Ocr::Ocr() |
| 171 |
{ |
| 172 |
} |
| 173 |
|
| 174 |
Ocr::~Ocr() |
| 175 |
{ |
| 176 |
clearTraning(); |
| 177 |
} |
| 178 |
|
| 179 |
void Ocr::trainFont( const QFont & font, QFontDatabase::WritingSystem writingSystem ) |
| 180 |
{ |
| 181 |
// checks |
| 182 |
if ( writingSystem != QFontDatabase::Latin ) { |
| 183 |
qWarning( "Ocr::trainFont: unsupported writing system" ); |
| 184 |
return; |
| 185 |
} |
| 186 |
|
| 187 |
// add chars |
| 188 |
QList<QChar> chars; |
| 189 |
// symbols !"#$%&'()*+,-./ |
| 190 |
//for ( int i = 33; i <= 47; i++ ) |
| 191 |
// chars.append( QChar( i ) ); |
| 192 |
// numbers |
| 193 |
//for ( int i = 48; i <= 57; i++ ) |
| 194 |
// chars.append( QChar( i ) ); |
| 195 |
// symbols :;<=>?@ |
| 196 |
//for ( int i = 58; i <= 64; i++ ) |
| 197 |
// chars.append( QChar( i ) ); |
| 198 |
// uppercase letters |
| 199 |
for ( int i = 65; i <= 90; i++ ) |
| 200 |
chars.append( QChar( i ) ); |
| 201 |
// symbols [\]^_` |
| 202 |
//for ( int i = 91; i <= 96; i++ ) |
| 203 |
// chars.append( QChar( i ) ); |
| 204 |
// lowercase letters |
| 205 |
//for ( int i = 97; i <= 122; i++ ) |
| 206 |
// chars.append( QChar( i ) ); |
| 207 |
// symbols {|}~ |
| 208 |
//for ( int i = 123; i <= 126; i++ ) |
| 209 |
// chars.append( QChar( i ) ); |
| 210 |
|
| 211 |
// generate glyphs for Latin1 and train the system |
| 212 |
QFontMetrics metrics( font ); |
| 213 |
foreach ( const QChar & character, chars ) { |
| 214 |
QRect rect = metrics.boundingRect( character ).adjusted( -5, -5, 5, 5 ); |
| 215 |
if ( !rect.isValid() ) { |
| 216 |
qWarning( "Ocr::trainFont: invalid rect for font" ); |
| 217 |
continue; |
| 218 |
} |
| 219 |
|
| 220 |
// draw glyph |
| 221 |
QImage image( rect.width(), rect.height(), QImage::Format_ARGB32 ); |
| 222 |
image.fill( 0xFFFFFFFF ); |
| 223 |
QPainter imgPainter( &image ); |
| 224 |
imgPainter.setFont( font ); |
| 225 |
imgPainter.setPen( Qt::black ); |
| 226 |
imgPainter.drawText( 0, 0, rect.width(), rect.height(), Qt::AlignCenter, character ); |
| 227 |
imgPainter.end(); |
| 228 |
|
| 229 |
// recognize glyph |
| 230 |
trainGlyph( image, character ); |
| 231 |
} |
| 232 |
} |
| 233 |
|
| 234 |
void Ocr::trainGlyph( const QImage & sourceImage, const QChar & character ) |
| 235 |
{ |
| 236 |
#if 0 |
| 237 |
QLabel * label = new QLabel(); |
| 238 |
label->setPixmap( QPixmap::fromImage( sourceImage ) ); |
| 239 |
label->setWindowTitle( character ); |
| 240 |
label->show(); |
| 241 |
#endif |
| 242 |
|
| 243 |
// checks |
| 244 |
if ( sourceImage.isNull() || sourceImage.width() < 4 || sourceImage.height() < 4 || character.isNull() ) { |
| 245 |
qWarning( "Ocr::trainGlyph: invalid input" ); |
| 246 |
return; |
| 247 |
} |
| 248 |
|
| 249 |
// prepare image for processing |
| 250 |
QImage image = trimImage( sourceImage ); |
| 251 |
|
| 252 |
// append glyph description |
| 253 |
OcrGlyph * glyph = new OcrGlyph(); |
| 254 |
glyph->character = character; |
| 255 |
glyph->image = image; |
| 256 |
glyph->width = image.width(); |
| 257 |
glyph->height = image.height(); |
| 258 |
glyph->ratio = (float)image.width() / (float)image.height(); |
| 259 |
m_glyphs.append( glyph ); |
| 260 |
} |
| 261 |
|
| 262 |
void Ocr::clearTraning() |
| 263 |
{ |
| 264 |
qDeleteAll( m_glyphs ); |
| 265 |
m_glyphs.clear(); |
| 266 |
} |
| 267 |
|
| 268 |
OcrResult Ocr::recognizeGlyph( const QImage & sourceImage, const QRect & __rect ) const |
| 269 |
{ |
| 270 |
// checks |
| 271 |
OcrResult result; |
| 272 |
result.character = '?'; // FIXME: REMOVE THIS |
| 273 |
result.confidence = 0.0; |
| 274 |
if ( sourceImage.isNull() ) |
| 275 |
return result; |
| 276 |
|
| 277 |
// prepare image for processing |
| 278 |
QRect rect = __rect.isValid() ? __rect : QRect( 0, 0, sourceImage.width(), sourceImage.height() ); |
| 279 |
QImage image = trimImage( sourceImage.copy( rect ) ); |
| 280 |
|
| 281 |
// match all glyphs |
| 282 |
foreach ( OcrGlyph * glyph, m_glyphs ) { |
| 283 |
OcrResult comparison = compareGlyph( image, glyph ); |
| 284 |
if ( comparison.confidence > result.confidence ) { |
| 285 |
result.character = comparison.character; |
| 286 |
result.confidence = comparison.confidence; |
| 287 |
} |
| 288 |
} |
| 289 |
|
| 290 |
// done! |
| 291 |
return result; |
| 292 |
} |