Updated OCR training utility to work better with Tesseract

src/misc_utilities/prepcharsfortraining.cpp

@@ -57,25 +57,45 @@ int main( int argc, const char** argv )
 
   if (DirectoryExists(inDir.c_str()))
   {
-    const int X_OFFSET = 10;
+    const int CHAR_PADDING_HORIZONTAL = 0;
-    const int Y_OFFSET = 10;
+    const int CHAR_PADDING_VERTICAL = 0;
+    
+    const int X_OFFSET = 5;
+    const int Y_OFFSET = 5;
 
     const int PAGE_MARGIN_X = 70;
     const int PAGE_MARGIN_Y = 70;
     const int HORIZONTAL_RESOLUTION = 3500;
+    const int MAX_VERTICAL_RESOLUTION = 6000; // Maximum vertical size before chopping into additional pages.
 
-    const int TILE_WIDTH = 55;
+    const int TILE_WIDTH = 25;
+    const int TILE_HEIGHT = 60;
     const int CHAR_HORIZ_OFFSET = 40;
-    const int TILE_HEIGHT = 70;
     const int CHAR_VERT_OFFSET = 48;
 
+    const int FIXED_CHAR_HEIGHT = 40; // RESIZE all characters to this height
+    
     vector<string> files = getFilesInDir(inDir.c_str());
 
     sort( files.begin(), files.end(), stringCompare );
-
+    
+    for (int i = 0; i< files.size(); i++)
+    {
+      if (hasEnding(files[i], ".png") || hasEnding(files[i], ".jpg"))
+      {
+	
+      }
+      else
+      {
+	std::cerr << "Non-image file detected in this directory.  This must be removed first" << std::endl;
+	return 1;
+      }
+    }
+    
+    
     int tiles_per_row = ((float) (HORIZONTAL_RESOLUTION - (PAGE_MARGIN_X * 2))) / ((float) TILE_WIDTH);
     int lines = files.size() / (tiles_per_row);
-    int vertical_resolution = (lines * TILE_HEIGHT) + (PAGE_MARGIN_Y * 2) ;
+    int vertical_resolution = (lines * TILE_HEIGHT) + (PAGE_MARGIN_Y * 3) ;
     cout << tiles_per_row <<   " : " << vertical_resolution << endl;
 
     Mat bigTif = Mat::zeros(Size(HORIZONTAL_RESOLUTION, vertical_resolution), CV_8U);
@@ -100,6 +120,8 @@ int main( int argc, const char** argv )
         char charcode = files[i][0];
 
         Mat characterImg = imread(fullpath);
+
+	
         Mat charImgCopy = Mat::zeros(Size(150, 150), characterImg.type());
         bitwise_not(charImgCopy, charImgCopy);
 
@@ -112,14 +134,34 @@ int main( int argc, const char** argv )
         //imshow("copy", charImgCopy);
         findContours(charImgCopy, contours, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_SIMPLE);
 
-        Rect tallestRect(0, 0, 0, 0);
+	float minHeightPercent = 0.35;
+	int minHeight = (int) (((float) characterImg.rows) * minHeightPercent);
+
+	vector<Rect> tallEnoughRects;
         for (int c = 0; c < contours.size(); c++)
         {
           Rect tmpRect = boundingRect(contours[c]);
-          if (tmpRect.height > tallestRect.height)
+          if (tmpRect.height > minHeight)
-            tallestRect = tmpRect;
+            tallEnoughRects.push_back( tmpRect );
         }
 
+	int xMin = 9999999, xMax = 0, yMin = 9999999, yMax = 0;
+        // Combine all the "tall enough" rectangles into one super rectangle
+        for (int r = 0; r < tallEnoughRects.size(); r++)
+	{
+	  if (tallEnoughRects[r].x < xMin)
+	    xMin = tallEnoughRects[r].x;
+	  if (tallEnoughRects[r].y < yMin)
+	    yMin = tallEnoughRects[r].y;
+	  if (tallEnoughRects[r].x + tallEnoughRects[r].width > xMax)
+	    xMax = tallEnoughRects[r].x + tallEnoughRects[r].width;
+	  if (tallEnoughRects[r].y + tallEnoughRects[r].height > yMax)
+	    yMax = tallEnoughRects[r].y + tallEnoughRects[r].height;
+	}
+	
+        Rect tallestRect(xMin, yMin, xMax - xMin, yMax - yMin);
+	
+	
         //cout << tallestRect.x << ":" << tallestRect.y << " -- " << tallestRect.width << ":" << tallestRect.height << endl;
 
         Rect cropRect(0, tallestRect.y - Y_OFFSET, tallestRect.width, tallestRect.height);
@@ -128,16 +170,16 @@ int main( int argc, const char** argv )
         Mat cropped(characterImg, cropRect);
         cvtColor(cropped, cropped, CV_BGR2GRAY);
 
-        Rect destinationRect(xPos + (CHAR_HORIZ_OFFSET - tallestRect.width), yPos + (CHAR_VERT_OFFSET - tallestRect.height), tallestRect.width, tallestRect.height);
+        Rect destinationRect(xPos + (CHAR_HORIZ_OFFSET - TILE_WIDTH), yPos + (CHAR_VERT_OFFSET - TILE_HEIGHT + (TILE_HEIGHT - tallestRect.height)), tallestRect.width, tallestRect.height);
 
         //cout << "1" << endl;
 
         cropped.copyTo(bigTif(destinationRect));
 
-        int x1= destinationRect.x - 2;
+        int x1 = destinationRect.x - CHAR_PADDING_HORIZONTAL;
-        int y1 = (vertical_resolution - destinationRect.y - destinationRect.height) - 2;
+        int y1 = (vertical_resolution - destinationRect.y - destinationRect.height) - CHAR_PADDING_VERTICAL;
-        int x2 = (destinationRect.x + destinationRect.width) + 2;
+        int x2 = (destinationRect.x + destinationRect.width) + CHAR_PADDING_HORIZONTAL;
-        int y2 = (vertical_resolution - destinationRect.y) + 2;
+        int y2 = (vertical_resolution - destinationRect.y) + CHAR_PADDING_VERTICAL;
         //0 70 5602 85 5636 0
         boxFileOut << charcode << " " << x1 << " " << y1 << " ";
         boxFileOut << x2 << " " << y2 ;