diff --git a/videocr/video.py b/videocr/video.py index b592dde..dbf2356 100644 --- a/videocr/video.py +++ b/videocr/video.py @@ -62,15 +62,14 @@ class Video: if not self.use_fullframe: # only use bottom half of the frame by default img = img[self.height // 2:, :] - # dilate and resize - img=cv2.resize(cv2.dilate(img, np.ones((2, 2), np.uint8)), self.resize_dim, interpolation=cv2.INTER_AREA) - - # mask to filter out non gray-like pixels/pixels that are not bright enough - hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV) - color_mask = cv2.inRange(hsv, (0, 0, 190), (179, 20, 255)) - - # apply mask, inverse image so it's black text on white background, add borders to top and bottom - img = cv2.copyMakeBorder(cv2.bitwise_not(cv2.bitwise_and(img, img, mask=color_mask)), 10, 10, 0, 0, cv2.BORDER_CONSTANT, None, (255,255,255)) + img = cv2.dilate(img, np.ones((2, 2), np.uint8)) + _, img = cv2.threshold(img, 215, 255, cv2.THRESH_BINARY) + color_mask = cv2.inRange(img, (255, 255, 255), (255, 255, 255)) + img = cv2.bitwise_and(img, img, mask=color_mask) + img = cv2.erode(img, np.ones((2, 2), np.uint8)) + img = cv2.bitwise_not(img) + img = cv2.resize(img, self.resize_dim, interpolation=cv2.INTER_AREA) + img = cv2.copyMakeBorder(img, 20, 20, 0, 0, cv2.BORDER_CONSTANT, None, (255,255,255)) config = '--tessdata-dir "{}" --psm 7 -c preserve_interword_spaces=1'.format(constants.TESSDATA_DIR) try: return pytesseract.image_to_data(img, lang=self.lang, config=config)