I have an image of a steak, and my goal is to calculate its marbling by converting the pixels to red for meat and white for fat. However, I have problems with glare in the image, which hinder the calculation. My code helps to improve it somewhat, but I would like to improve it even further. Here is an example, I have this image below, with glare:

my code is this:

import cv2
import numpy as np
import matplotlib.pyplot as plt

# Load the image
img = cv2.imread("./img.jpg")

# Convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# Apply a Gaussian filter to smooth the image
blurred = cv2.GaussianBlur(gray, (15, 15), 0)

# Calculate the difference between the original image and the smoothed one to highlight brightness
diff = cv2.absdiff(gray, blurred)

# Apply a threshold to highlight intensely bright areas
_, mask = cv2.threshold(diff, 35, 255, cv2.THRESH_BINARY)

# Invert the mask to get the dark areas (which might be glare)
mask = 255 - mask

# Apply the mask to the original image to remove glare
result = cv2.bitwise_and(img, img, mask=mask)

cv2.imwrite('marmoreio_red_5.jpg', result)

ret, data_1 = cv2.threshold(result, 127, 240, cv2.THRESH_BINARY)  # Standard LG ## Various 127
cv2.imwrite("filter_1.jpeg", data_1)

# Transform non-black and non-red pixels into white
for y in range(data_1.shape[0]):
    for x in range(data_1.shape[1]):
        pixel = data_1[y, x]

        # Define what is considered 'red' and 'black'
        # (adjust these thresholds as necessary)
        isRed = pixel[2] > 150 and pixel[1] < 50 and pixel[0] < 50
        isBlack = pixel[2] < 50 and pixel[1] < 50 and pixel[0] < 50

        if not isRed and not isBlack:
            # If not red or black, set to white
            pixel[0] = 255  # B
            pixel[1] = 255  # G
            pixel[2] = 255  # R
        elif isRed:
            # If it is red, standardize to the same shade of red
            pixel[0] = 0  # B
            pixel[1] = 0  # G
            # Keep the red component as it is
        elif isBlack:
            # If it is black, transform to red
            pixel[0] = 0  # B
            pixel[1] = 0  # G
            pixel[2] = 255  # R

cv2.imwrite("filter_2.jpeg", data_1)

The code helps to improve the glare, but there’s still a lot of noise. Does anyone have a tip on what I can do to improve it? Here’s the final processed image:”