Migrate to OpenCV 5

GetLargestContour.py

@@ -62,8 +62,8 @@
     cv2.drawContours(output, hull, -1, (0,255,0), 2)
     largestcontourarea = cv2.contourArea(hull[0])
 
-    for p in hull[0]:
-        polygon.append({'x':p[0][0], 'y':p[0][1]})
+    for p in np.atleast_2d(np.squeeze(hull[0])):
+        polygon.append({'x':p[0], 'y':p[1]})
     #cv2.rectangle(output,(x,y),(x+w,y+h),(0,255,0),2)
 
     sd = CropPolygonsToSingleImage()

ImageProcess/CalculatePhenotypeSift.py

@@ -27,7 +27,7 @@
 result_file_lines = []
 for image in images:
     img = cv2.imread(image)
-    descriptor = cv2.xfeatures2d.SIFT_create()
+    descriptor = cv2.SIFT_create()
     (kps, features) = descriptor.detectAndCompute(img, None)
     print(features)
     result_file_lines.append([len(kps)])

ImageProcess/MatchAndAlignImages.py

@@ -34,7 +34,7 @@
 img2 = cv2.imread(input_image1, cv2.IMREAD_UNCHANGED)
 
 # orb = cv2.ORB_create(MAX_FEATURES)
-orb = cv2.xfeatures2d.SIFT_create(MAX_FEATURES)
+orb = cv2.SIFT_create(MAX_FEATURES)
 keypoints1, descriptors1 = orb.detectAndCompute(img1, None)
 keypoints2, descriptors2 = orb.detectAndCompute(img2, None)
 

ImageProcess/MergeChannels.py

@@ -21,7 +21,7 @@ def align_images(moving, fixed_im):
             moving_im = b
 
     # Initiate SIFT detector
-    sift = cv2.xfeatures2d.SIFT_create()
+    sift = cv2.SIFT_create()
     print("GET SIFT")
 
     # find the keypoints and descriptors with SIFT

ImageProcess/RemoveBackground.py

@@ -23,7 +23,8 @@
 
 src = cv2.imread(input_image, cv2.IMREAD_GRAYSCALE)
 
-th, dst = cv2.threshold(src, int(float(lower_thresh)), int(float(upper_thresh)), cv2.THRESH_TOZERO)
+th, dst = cv2.threshold(src, int(float(lower_thresh)), 0, cv2.THRESH_TOZERO) # zero out pixels darker than lower_thresh
+th, dst = cv2.threshold(dst, int(float(upper_thresh)), 0, cv2.THRESH_TOZERO_INV) # zero out pixels brighter than upper_thresh
 
 #cv2.imshow("Result", dst)
 cv2.imwrite(outfile_path, dst)

ImageProcess/RemoveBackgroundPercentage.py

@@ -29,6 +29,9 @@
 img_shape = img.shape
 
 if len(img_shape) == 3:
+    if img_shape[2] == 4:
+        img = cv2.cvtColor(img, cv2.COLOR_BGRA2BGR) # drop the alpha channel so it is not thresholded along with the color channels
+        img_shape = img.shape
     if img_shape[2] == 3:
         b,g,r = cv2.split(img)
         if image_band_index is not None:
@@ -41,7 +44,7 @@
                 img = r
 
 
-hist = cv2.calcHist([img],[0],None,histSize,[0,255])
+hist = cv2.calcHist([img],[0],None,histSize,[0,256]).flatten()
 
 #print(hist)
 #print(img.shape)
@@ -51,16 +54,16 @@
 drone_imagery_remove_background_lower_percentage_threshold = 0
 drone_imagery_remove_background_upper_percentage_threshold = 0
 for i in range(0, histSize[0]):
-    binVal = hist[i][0]
+    binVal = hist[i]
     summing = summing + binVal
     percentage = summing / total_pixels
     if percentage >= lower_percentage:
         drone_imagery_remove_background_lower_percentage_threshold = i
         break
 
-summing = 0;
+summing = 0
 for i in range(0, histSize[0]):
-    binVal = hist[i][0]
+    binVal = hist[i]
     summing = summing + binVal
     percentage = summing / total_pixels
     if percentage >= 1-upper_percentage:
@@ -69,7 +72,8 @@
 
 lower_thresh = int(float(drone_imagery_remove_background_lower_percentage_threshold))
 upper_thresh = int(float(drone_imagery_remove_background_upper_percentage_threshold))
-th, dst = cv2.threshold(img, lower_thresh, upper_thresh, cv2.THRESH_TOZERO)
+th, dst = cv2.threshold(img, lower_thresh, 0, cv2.THRESH_TOZERO) # zero out the darkest lower_percentage of pixels
+th, dst = cv2.threshold(dst, upper_thresh, 0, cv2.THRESH_TOZERO_INV) # zero out the brightest upper_percentage of pixels
 
 cv2.imwrite(outfile_path, dst)
 #cv2.waitKey(0)

ImageStitching/Panorama/Panorama.py

@@ -82,7 +82,7 @@ def detectAndDescribe(self, image):
 
         # Only works in OpenCV 3.X
         # detect and extract features from the image
-        descriptor = cv2.xfeatures2d.SIFT_create()
+        descriptor = cv2.SIFT_create()
         (kps, features) = descriptor.detectAndCompute(image, None)
 
         # convert the keypoints from KeyPoint objects to NumPy arrays