Search code examples
pythonocrtypeerrortesseractpytesser

"ValueError: cannot filter palette images" during Pytesseract Conversion

Having trouble with this error code regarding the following code for Pytesseract. (Python 3.6.1, Mac OSX)

import pytesseract import requests from PIL import Image from PIL import ImageFilter from io import StringIO, BytesIO

def process_image(url):
    image = _get_image(url)
    image.filter(ImageFilter.SHARPEN)
    return pytesseract.image_to_string(image)


def _get_image(url):
    r = requests.get(url)
    s = BytesIO(r.content)
    img = Image.open(s)
    return img

process_image("https://www.prepressure.com/images/fonts_sample_ocra_medium.png")

Error:

/usr/local/Cellar/python3/3.6.0_1/Frameworks/Python.framework/Versions/3.6/bin/python3.6 /Users/g/pyfo/reddit/ocr.py
Traceback (most recent call last):
  File "/Users/g/pyfo/reddit/ocr.py", line 20, in <module>
    process_image("https://www.prepressure.com/images/fonts_sample_ocra_medium.png")
  File "/Users/g/pyfo/reddit/ocr.py", line 10, in process_image
    image.filter(ImageFilter.SHARPEN)
  File "/usr/local/lib/python3.6/site-packages/PIL/Image.py", line 1094, in filter
    return self._new(filter.filter(self.im))
  File "/usr/local/lib/python3.6/site-packages/PIL/ImageFilter.py", line 53, in filter
    raise ValueError("cannot filter palette images")
ValueError: cannot filter palette images

Process finished with exit code 1

Seems simple enough, but is not working. Any help would be greatly appreciated.

Solution

  • The image you have is a pallet-based image. You need to convert it to a full RGB image in order to use the PIL filters.

    import pytesseract 
import requests 
from PIL import Image, ImageFilter 
from io import StringIO, BytesIO

def process_image(url):
    image = _get_image(url)
    image = image.convert('RGB')
    image = image.filter(ImageFilter.SHARPEN)
    return pytesseract.image_to_string(image)


def _get_image(url):
    r = requests.get(url)
    s = BytesIO(r.content)
    img = Image.open(s)
    return img

process_image("https://www.prepressure.com/images/fonts_sample_ocra_medium.png")

    You should also note that the the .convert() and .filter() methods return a copy of the image, they don't change the existing image object. You need to assign the return value to a variable as shown in the code above.

    NOTE: I don't have pytesseract, so I can't check the last line of process_image().