UnicodeDecodeError with Tesseract OCR in Python

7
votes

Iam trying to extract text from an image file using Tesseract OCR in Python but I'am facing an Error that i can figure out how to deal with it. all my environment is good as i tested some sample image with the ocr in python!

here is the code

from PIL import Image
import pytesseract
strs = pytesseract.image_to_string(Image.open('binarized_image.png'))

print (strs)

the follow is the error I get from eclipse console

strs = pytesseract.image_to_string(Image.open('binarized_body.png'))
  File "C:\Python35x64\lib\site-packages\pytesseract\pytesseract.py", line 167, in image_to_string
    return f.read().strip()
  File "C:\Python35x64\lib\encodings\cp1252.py", line 23, in decode
    return codecs.charmap_decode(input,self.errors,decoding_table)[0]
UnicodeDecodeError: 'charmap' codec can't decode byte 0x9d in position 20: character maps to <undefined>

Iam using python 3.5 x64 on Windows10

2
pythontesseractpython-tesseract
This reminds me of something I've encountered in the past; I don't know if it's exactly the same issue though. The fact that you're on Windows tipped me off - Python in CMD on windows seems to have a strange default code page. Have you tried hacking around at sys.setdefaultencoding to see if that helps you diagnose the problem? (I'd probably avoid keeping that hack around in production code if you can help it though.) – Benjamin Hodgson
Possible duplicate of Pytesseract: UnicodeDecodeError: 'charmap' codec can't decode byte – Sreeragh A R

2 Answers

8
votes

The problem is that python is trying to use the console's encoding (CP1252) instead of what it's meant to use (UTF-8). PyTesseract has found a unicode character and is now trying to translate it into CP1252, which it can't do. On another platform you won't encounter this error because it will get to use UTF-8.

You can try using a different function (possibly one that returns bytes instead of str so you won't have to worry about encoding). You could change the default encoding of python as mentioned in one of the comments, although that will cause problems when you go to try and print the string on the windows console. Or, and this is my recommended solution, you could download Cygwin and run python on that to get a clean UTF-8 output.

If you want a quick and dirty solution that won't break anything (yet), here's a way that you might consider:

import builtins

original_open = open
def bin_open(filename, mode='rb'):       # note, the default mode now opens in binary
    return original_open(filename, mode)

from PIL import Image
import pytesseract

img = Image.open('binarized_image.png')

try:
    builtins.open = bin_open
    bts = pytesseract.image_to_string(img)
finally:
    builtins.open = original_open

print(str(bts, 'cp1252', 'ignore'))
3
votes

I've had the same problem as you but I had to save the output of pytesseract to a file. So, I created a function for ocr with pytesseract and when saving to a file added parameter encoding='utf-8' so my function now looks like this:

def image_ocr(image_path, output_txt_file_name):
  image_text = pytesseract.image_to_string(image_path, lang='eng+ces', config='--psm 1')
  with open(output_txt_file_name, 'w+', encoding='utf-8') as f:
    f.write(image_text)

I hope this helps someone :)