Merge pull request #126 from BamX/default-resolution

Added default source resolution and clamping to min/max values.

Author: kevincon

TesseractOCR/G8Tesseract.h

@@ -11,6 +11,21 @@
 #import "G8Constants.h"
 #import "G8TesseractDelegate.h"
 
+/**
+ *  Default value of `sourceResolution` property.
+ */
+extern NSInteger const kG8DefaultResolution;
+
+/**
+ *  Minimum alowed value of `sourceResolution` property.
+ */
+extern NSInteger const kG8MinCredibleResolution;
+
+/**
+ *  Maximum alowed value of `sourceResolution` property.
+ */
+extern NSInteger const kG8MaxCredibleResolution;
+
 /**
  *  `G8Tesseract` encapsulates the Tesseract library and provides properties
  *  and methods for performing text recognition and analysis on a target image.
@@ -95,7 +110,10 @@
 
 /**
  *  The resolution of the source image in pixels per inch so font size
- *  information can be calculated in results. Change it after setting image.
+ *  information can be calculated in results. It should be from 70 to
+ *  2400 or it will be clamped.
+ *
+ *  @default Default value is 72
  */
 @property (nonatomic, assign) NSInteger sourceResolution;
 

TesseractOCR/G8Tesseract.mm

@@ -22,6 +22,10 @@
 #import "genericvector.h"
 #import "strngs.h"
 
+NSInteger const kG8DefaultResolution = 72;
+NSInteger const kG8MinCredibleResolution = 70;
+NSInteger const kG8MaxCredibleResolution = 2400;
+
 namespace tesseract {
     class TessBaseAPI;
 };
@@ -110,7 +114,7 @@ - (id)initWithLanguage:(NSString *)language
         _engineMode = engineMode;
         _pageSegmentationMode = G8PageSegmentationModeSingleBlock;
         _variables = [NSMutableDictionary dictionary];
-        _sourceResolution = 300;
+        _sourceResolution = kG8DefaultResolution;
         _rect = CGRectZero;
 
         _monitor = new ETEXT_DESC();
@@ -393,7 +397,6 @@ - (void)setImage:(UIImage *)image
         pixDestroy(&pix);
 
         _image = image;
-        _sourceResolution = 300;
         _rect = (CGRect){CGPointZero, self.imageSize};
 
         [self resetFlags];
@@ -439,6 +442,15 @@ - (void)setRect:(CGRect)rect
 - (void)setSourceResolution:(NSInteger)sourceResolution
 {
     if (_sourceResolution != sourceResolution) {
+        if (sourceResolution > kG8MaxCredibleResolution) {
+            NSLog(@"Source resolution is too big: %ld > %ld", (long)sourceResolution, (long)kG8MaxCredibleResolution);
+            sourceResolution = kG8MaxCredibleResolution;
+        }
+        else if (sourceResolution < kG8MinCredibleResolution) {
+            NSLog(@"Source resolution is too small: %ld < %ld", (long)sourceResolution, (long)kG8MinCredibleResolution);
+            sourceResolution = kG8MinCredibleResolution;
+        }
+        
         _sourceResolution = sourceResolution;
 
         _tesseract->SetSourceResolution((int)sourceResolution);
@@ -801,7 +813,7 @@ - (Pix *)pixForImage:(UIImage *)image
         default:
             NSLog(@"Cannot convert image to Pix with bpp = %d", bpp);
     }
-    pixSetYRes(pix, 300);
+    pixSetYRes(pix, (l_int32)self.sourceResolution);
 
     CFRelease(imageData);
 

TestsProject/TestsProject/well_scaned_page.hOCR

@@ -5,7 +5,7 @@
      </span>
     </p>
    </div>
-   <div class='ocr_carea' id='block_1_2' title="bbox 7 79 805 1022">
+   <div class='ocr_carea' id='block_1_2' title="bbox 7 79 806 1022">
     <p class='ocr_par' dir='ltr' id='par_1_2' title="bbox 9 79 804 251">
      <span class='ocr_line' id='line_1_2' title="bbox 53 79 804 102; baseline 0.003 -6"><span class='ocrx_word' id='word_1_2' title='bbox 53 79 95 96; x_wconf 85' lang='eng' dir='ltr'>This</span> <span class='ocrx_word' id='word_1_3' title='bbox 101 80 169 96; x_wconf 79' lang='eng' dir='ltr'><strong>volume</strong></span> <span class='ocrx_word' id='word_1_4' title='bbox 175 80 189 96; x_wconf 90' lang='eng' dir='ltr'><strong>is</strong></span> <span class='ocrx_word' id='word_1_5' title='bbox 196 80 224 97; x_wconf 89' lang='eng' dir='ltr'><strong>the</strong></span> <span class='ocrx_word' id='word_1_6' title='bbox 230 81 283 97; x_wconf 96' lang='eng' dir='ltr'><strong>result</strong></span> <span class='ocrx_word' id='word_1_7' title='bbox 289 81 307 97; x_wconf 79' lang='eng' dir='ltr'><strong>of</strong></span> <span class='ocrx_word' id='word_1_8' title='bbox 313 81 342 97; x_wconf 88' lang='eng' dir='ltr'><strong>the</strong></span> <span class='ocrx_word' id='word_1_9' title='bbox 348 81 455 101; x_wconf 65' lang='eng' dir='ltr'><strong>cooperative</strong></span> <span class='ocrx_word' id='word_1_10' title='bbox 462 81 513 97; x_wconf 61' lang='eng' dir='ltr'><strong>efl’ort</strong></span> <span class='ocrx_word' id='word_1_11' title='bbox 518 81 536 97; x_wconf 90' lang='eng' dir='ltr'><strong>of</strong></span> <span class='ocrx_word' id='word_1_12' title='bbox 541 87 594 102; x_wconf 79' lang='eng' dir='ltr'><strong>many</strong></span> <span class='ocrx_word' id='word_1_13' title='bbox 600 87 669 102; x_wconf 75' lang='eng' dir='ltr'><strong>persons</strong></span> <span class='ocrx_word' id='word_1_14' title='bbox 676 82 709 98; x_wconf 88' lang='eng' dir='ltr'><strong>and</strong></span> <span class='ocrx_word' id='word_1_15' title='bbox 716 87 726 98; x_wconf 87' lang='eng' dir='ltr'><strong>a</strong></span> <span class='ocrx_word' id='word_1_16' title='bbox 732 84 804 98; x_wconf 82' lang='eng' dir='ltr'><strong>number</strong></span> 
      </span>

TestsProject/TestsProjectTests/G8RecognitionTestsHelper.h

@@ -31,6 +31,7 @@ typedef NS_ENUM(NSUInteger, G8CustomPreprocessing) {
 @property (nonatomic, assign) CGSize boundingSizeForResizing;
 @property (nonatomic, assign) G8CustomPreprocessing customPreprocessingType;
 
+- (void)setupTesseract;
 - (void)recognizeImage;
 - (void)recognizeImageUsingOperation;
 - (UIImage *)thresholdedImageForImage:(UIImage *)sourceImage;

TestsProject/TestsProjectTests/G8RecognitionTestsHelper.m

@@ -50,6 +50,10 @@ - (void)waitTimeLmit:(NSTimeInterval)maximumWait whileTrue:(BOOL (^)())shouldKee
 
 - (void)setupTesseract
 {
+    if (self.tesseract == nil) {
+        self.tesseract = [[G8Tesseract alloc] init];
+    }
+    
     self.tesseract.delegate = self;
 
     self.tesseract.language = kG8Languages;

TestsProject/TestsProjectTests/RecognitionTests.m

@@ -125,6 +125,20 @@
         [[theValue(block.level) should] equal:theValue(G8PageIteratorLevelWord)];
     });
 
+    it(@"Should clamp source resolution", ^{
+        [helper setupTesseract];
+
+        [[theValue(helper.tesseract.sourceResolution) should] equal:theValue(kG8DefaultResolution)];
+
+        helper.tesseract.sourceResolution = 50;
+        [[theValue(helper.tesseract.sourceResolution) should] beInTheIntervalFrom:theValue(kG8MinCredibleResolution)
+                                                                               to:theValue(kG8MaxCredibleResolution)];
+
+        helper.tesseract.sourceResolution = 3000;
+        [[theValue(helper.tesseract.sourceResolution) should] beInTheIntervalFrom:theValue(kG8MinCredibleResolution)
+                                                                               to:theValue(kG8MaxCredibleResolution)];
+    });
+
     it(@"Should draw blocks on image", ^{
         [helper recognizeImage];