How to improve the accuracy of reading text from an image?

Question

Python

from PIL import Image
from pytesseract import image_to_string
import requests
from io import BytesIO
import re

url = 'https://pbs.twimg.com/media/Dh-ZHUPX4AAQwG-.jpg:large'

response = requests.get(url)

img = Image.open(BytesIO(response.content))

text_from_image = image_to_string(img)

print(text_from_image)

What image manipulations can be performed to improve the accuracy of reading text from an image?
At the moment, the reading accuracy does not exceed 40%.

Answer 1

sharpen increase

Answer 2

Crop everything superfluous (in this case, the image of a weapon), increase sharpness, remove the background (gray) and convert the resulting image to a black and white image. After such manipulations, you will get a white background and black text on it. His tesseract recognizes much better.