Image preprocessing for improving ocr

preprocessing.py

import os
from typing import Tuple

import cv2
import numpy as np


def image_binarization(image: np.ndarray) -> np.ndarray:
    if len(image.shape) > 2:  # convert to gray scale if not in gray scale
        image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    out_binary = cv2.threshold(noise_removal(image), 0, 255, cv2.THRESH_OTSU)[1]
    return out_binary


def noise_removal(image: np.ndarray) -> np.ndarray:
    if len(image.shape) > 2:  # convert to gray scale if not in gray scale
        image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    se = cv2.getStructuringElement(cv2.MORPH_RECT, (8, 8))
    bg = cv2.morphologyEx(image, cv2.MORPH_DILATE, se)
    out_gray = cv2.divide(image, bg, scale=255)
    return out_gray


def denoising(image: np.ndarray):
    return cv2.fastNlMeansDenoisingColored(image, None, 10, 10, 7, 21)


def unsharp_mask(
    image: np.ndarray,
    kernel_size: Tuple[int, int] = (5, 5),
    sigma: float = 1.0,
    amount: float = 1.0,
    threshold: int = 0,
):
    blurred = cv2.GaussianBlur(image, kernel_size, sigma)
    sharpened = float(amount + 1) * image - float(amount) * blurred
    sharpened = np.maximum(sharpened, np.zeros(sharpened.shape))
    sharpened = np.minimum(sharpened, 255 * np.ones(sharpened.shape))
    sharpened = sharpened.round().astype(np.uint8)
    if threshold > 0:
        low_contrast_mask = np.absolute(image - blurred) < threshold
        np.copyto(sharpened, image, where=low_contrast_mask)
    return sharpened


image = cv2.imread(os.path.join("examples", "image.png")
im_bin = image_binarization(image)
out_gray = noise_removal(image)


speckle = cv2.imread(os.path.join("examples", "speckle.png"))
speckle = denoising(speckle)

blurry = cv2.imread(os.path.join("examples", "blurry.png"))
blurry = unsharp_mask(blurry)

cv2.imwrite("im_bin.png", im_bin)
cv2.imwrite("out_gray.png", out_gray)
cv2.imwrite("un_speckle.png", speckle)
cv2.imwrite("un_blurry.png", blurry)