Maybe this is what you mean.
import cv2 import numpy as np import pytesseract import imutils img = cv2.imread("srj8n.png") cv2.imshow("original", img) # turn into gray for next processing gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY_INV * cv2.THRESH_OTSU)[1] thresh = cv2.bitwise_not(thresh) # omit the underline kernel = np.ones((4, 4), np.uint8) erosion = cv2.erode(thresh, kernel, iterations=1) # dilate to make the line thicker kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (9, 12)) dilation = cv2.dilate(erosion, kernel, iterations=1) # find the contour cntrs = cv2.findContours(dilation, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) cntrs = cntrs[0] if len(cntrs) == 2 else cntrs[1] result = img.copy() for c in cntrs: # for each letter, create red rectangle x, y, w, h = cv2.boundingRect(c) cv2.rectangle(result, (x, y), (x + w, y + h), (0, 0, 255), 2) # prepare letter for OCR box = thresh[y:y + h - 2, x:x + w] box = cv2.bitwise_not(box) box = cv2.GaussianBlur(box, (3, 3), 0) # retreive the angle. For the meaning of angle, see below # https://namkeenman.wordpress.com/2015/12/18/open-cv-determine-angle-of-rotatedrect-minarearect/ rect = cv2.minAreaRect(c) angle = rect[2] # put angle below letter font = cv2.FONT_HERSHEY_SIMPLEX bottomLeftCornerOfText = (x, y+h+20) fontScale = 0.6 fontColor = (255, 0, 0) lineType = 2 cv2.putText(result, str(angle), bottomLeftCornerOfText, font, fontScale, fontColor, lineType) # do the OCR custom_config = r'-l eng --oem 3 --psm 10' text = pytesseract.image_to_string(box, config=custom_config) print("Detected :" + text + ", angle: " + str(angle)) cv2.imshow("result", result) cv2.waitKey(0) cv2.destroyAllWindows()
