Correct text-image orientation with Python/Tesseract/OpenCV

orient.py

import cv2
import pytesseract
import urllib
import numpy as np
import re

# Installs: https://www.learnopencv.com/deep-learning-based-text-recognition-ocr-using-tesseract-and-opencv/

if __name__ == '__main__':
     
  # Uncomment the line below to provide path to tesseract manually
  # pytesseract.pytesseract.tesseract_cmd = '/usr/bin/tesseract'
 
 
  # Read image from URL
  #   Taken from https://stackoverflow.com/questions/21061814/how-can-i-read-an-image-from-an-internet-url-in-python-cv2-scikit-image-and-mah
  # https://i.ibb.co/4mm9WvZ/book-rot.jpg
  # https://i.ibb.co/M7jwWR2/book.jpg
  # https://i.ibb.co/27bKNJ8/book-rot2.jpg
  resp = urllib.request.urlopen('https://i.ibb.co/27bKNJ8/book-rot2.jpg')
  image = np.asarray(bytearray(resp.read()), dtype="uint8")
  image = cv2.imdecode(image, cv2.IMREAD_COLOR) # Initially decode as color
  
  #  TAKEN FROM: https://www.pyimagesearch.com/2017/02/20/text-skew-correction-opencv-python/
  # convert the image to grayscale and flip the foreground
  # and background to ensure foreground is now "white" and
  # the background is "black"
  gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
  gray = cv2.bitwise_not(gray)
 
  rot_data = pytesseract.image_to_osd(image);
  print("[OSD] "+rot_data)
  rot = re.search('(?<=Rotate: )\d+', rot_data).group(0)

  angle = float(rot)
  if angle > 0:
    angle = 360 - angle
  print("[ANGLE] "+str(angle))
    
  # rotate the image to deskew it
  (h, w) = image.shape[:2]
  center = (w // 2, h // 2)
  M = cv2.getRotationMatrix2D(center, angle, 1.0)
  rotated = cv2.warpAffine(image, M, (w, h),
	flags=cv2.INTER_CUBIC, borderMode=cv2.BORDER_REPLICATE)	
	
  #  TODO: Rotated image can be saved here
  print(pytesseract.image_to_osd(rotated));
  print("[TEXT]")
  # Run tesseract OCR on image
  text = pytesseract.image_to_string(rotated, lang='eng', config="-psm 1")

  # Print recognized text
  print(text.encode(encoding='UTF-8'))
                

The code is very helpful. However I found one change that could really improve it.
Currently you are finding an angle and deciding whether the image should be rotated clockwise or anti clockwise. Then you are using cv2 functions to get the rotated matrix and warp affine. The problem is that OpenCV does not automatically allocate space for our entire rotated image to fit into the frame. As a result, if your image is of rectangular shape and rotated +90 or -90 then when you correct its orientation, half the image will be missing.

To avoid that you can use rotation using bound awareness. I am using the imutils function from https://www.pyimagesearch.com/2021/01/20/opencv-rotate-image/

import cv2
import pytesseract
import urllib
import numpy as np
import re
import imutils #added

resp = urllib.request.urlopen('https://i.ibb.co/27bKNJ8/book-rot2.jpg')
image = np.asarray(bytearray(resp.read()), dtype="uint8")
image = cv2.imdecode(image, cv2.IMREAD_COLOR) # Initially decode as color
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
gray = cv2.bitwise_not(gray)

rot_data = pytesseract.image_to_osd(image);
print("[OSD] "+rot_data)
rot = re.search('(?<=Rotate: )\d+', rot_data).group(0)

angle = float(rot)

# rotate the image to deskew it
rotated = imutils.rotate_bound(image, angle) #added


#  TODO: Rotated image can be saved here
print(pytesseract.image_to_osd(rotated));
print("[TEXT]")
# Run tesseract OCR on image
text = pytesseract.image_to_string(rotated, lang='eng', config="--psm 6")

# Print recognized text