Fix #52 Comparisons with empty masks

Author: Avasam

src/compare.py

@@ -16,10 +16,10 @@ def compare_histograms(source, capture):
 
     source_hist = cv2.calcHist([source], [0, 1, 2], None, [8, 8, 8], [0, 256, 0, 256, 0, 256])
     capture_hist = cv2.calcHist([capture], [0, 1, 2], None, [8, 8, 8], [0, 256, 0, 256, 0, 256])
-    
+
     cv2.normalize(source_hist, source_hist)
     cv2.normalize(capture_hist, capture_hist)
-    
+
     return 1 - cv2.compareHist(source_hist, capture_hist, cv2.HISTCMP_BHATTACHARYYA)
 
 def compare_histograms_masked(source, capture, mask):
@@ -34,10 +34,10 @@ def compare_histograms_masked(source, capture, mask):
     """
     source_hist = cv2.calcHist([source], [0, 1, 2], mask, [8, 8, 8], [0, 256, 0, 256, 0, 256])
     capture_hist = cv2.calcHist([capture], [0, 1, 2], mask, [8, 8, 8], [0, 256, 0, 256, 0, 256])
-    
+
     cv2.normalize(source_hist, source_hist)
     cv2.normalize(capture_hist, capture_hist)
-    
+
     return 1 - cv2.compareHist(source_hist, capture_hist, cv2.HISTCMP_BHATTACHARYYA)
 
 def compare_l2_norm(source, capture):
@@ -51,10 +51,10 @@ def compare_l2_norm(source, capture):
     """
 
     error = cv2.norm(source, capture, cv2.NORM_L2)
-    
+
     # The L2 Error is summed across all pixels, so this normalizes
     max_error = (source.size ** 0.5) * 255
-    
+
     return 1 - (error/max_error)
 
 def compare_l2_norm_masked(source, capture, mask):
@@ -73,6 +73,8 @@ def compare_l2_norm_masked(source, capture, mask):
     # The L2 Error is summed across all pixels, so this normalizes
     max_error = (3 * numpy.count_nonzero(mask) * 255 * 255) ** 0.5
 
+    if not max_error:
+        return 0
     return 1 - (error / max_error)
 
 def compare_template(source, capture):
@@ -117,7 +119,7 @@ def compare_phash(source, capture):
     """
     Compares the pHash of the two given images and returns the similarity between
     the two.
-    
+
     @param source: Image of any given shape as a numpy array
     @param capture: Image of any given shape as a numpy array
     @return: The similarity between the hashes of the image as a number 0 to 1.
@@ -135,7 +137,7 @@ def compare_phash_masked(source, capture, mask):
     """
     Compares the pHash of the two given images and returns the similarity between
     the two.
-    
+
     @param source: Image of any given shape as a numpy array
     @param capture: Image of any given shape as a numpy array
     @param mask: An image matching the dimensions of the source, but 1 channel grayscale
@@ -149,11 +151,27 @@ def compare_phash_masked(source, capture, mask):
     # the same
     source = cv2.bitwise_and(source, source, mask=mask)
     capture = cv2.bitwise_and(capture, capture, mask=mask)
-    
+
     source = Image.fromarray(source)
     capture = Image.fromarray(capture)
 
     source_hash = imagehash.phash(source)
     capture_hash = imagehash.phash(capture)
 
-    return 1 - ((source_hash - capture_hash)/64.0)
+    if not source_hash + capture_hash:
+        return 0
+    return 1 - ((source_hash - capture_hash) / 64.0)
+
+
+def checkIfImageHasTransparency(image):
+    # Check if there's a transparency channel (4th channel) and if at least one pixel is transparent (< 255)
+    if image.shape[2] != 4:
+        return False
+    mean = np.mean(image[:, :, 3])
+    if mean != 0:
+        # Non-transparent images code path is usually faster and simpler, so let's return that
+        return False
+        # TODO error message if all pixels are transparent
+        # (the image appears as all black in windows, so it's not obvious for the user what they did wrong)
+
+    return mean != 255