Red-eye removal in python using opencv

Question

i have a code : that kind of fixes the red-eye

the output is somewhat correct but it applying redeye over other parts of face including the eye. i want it to be just in the eyes.

i have tried out few online reference non worked in my case any assistance will be appreciated

output image : output

import cv2
import numpy as np

def fillHoles(mask):
    maskFloodFill = mask.copy()
    h, w = maskFloodFill.shape[:2]
    maskTemp = np.zeros((h+2, w+2), np.uint8)
    cv2.floodFill(maskFloodFill, maskTemp, (0, 0), 255)
    mask2 = cv2.bitwise_not(maskFloodFill)
    return mask2 | mask


def removeRedEye(img, red_eye_threshold):
    height, width = img.shape[:2]
    img_fixed = img.copy()

    face_cascade_name = "./haarcascade_face.xml"
    eyes_cascade_name = "./haarcascade_eye.xml"

    eyes_cascade = cv2.CascadeClassifier()

    assert eyes_cascade.load(eyes_cascade_name), "Eyes detection model loading failed."

    eyes = eyes_cascade.detectMultiScale(img, scaleFactor=1.1, minNeighbors=3, flags=0,
                                         minSize=(height//50, width//50), maxSize=(height//2, width//2))

    for (x, y, w, h) in eyes:
        eye = img[y:y+h, x:x+w]

        b, g, r = cv2.split(eye)

        mask = (r > red_eye_threshold) & (r > (b + g))

        mask = mask.astype(np.uint8)*255
        fillHoles(mask)

        mask = cv2.dilate(mask, None, iterations=3)
        mean = np.mean(eye[:, :, :2], axis=2)
        mean = mean[:, :, np.newaxis]

        eye_fixed = np.dstack((mean, mean, mean))

        img_fixed[y:y+h, x:x+w] = np.where(mask[:, :, np.newaxis], eye_fixed, img_fixed[y:y+h, x:x+w])

    return img_fixed

Need some quick fixes for it. Guide where am going wrong !

Can you please provide suitible test data, or add examples to your question, so we can understand the problem more clearly? — Christian Schaffner, Jun 23 '23 at 07:11
What is the intention of the comment "Eyes detection model loading failed"? — Markus, Jun 23 '23 at 07:26
the output has red patches along the eye area @ChristianSchaffner — Maximus, Jun 23 '23 at 07:36
It seems as if your model does not detect the area of your eye, but a rectangle containing the eye. As you can see the output of the detection is only a width and height as well as a position — Christian Schaffner, Jun 23 '23 at 07:45
to detect pixels of a red eye, use the solution from this post: https://stackoverflow.com/questions/64816949/confusion-with-red-eye-detection — Christian Schaffner, Jun 23 '23 at 07:47
am not quite clear. can you provide me an code for removing red-eye ? — Maximus, Jun 23 '23 at 07:57
You could start with the code used in this question: https://stackoverflow.com/questions/57067036/can-we-use-haar-classifier-to-detect-iris-of-an-eye-as-done-for-face-and-eyes or search for tutorials about iris detection. — Markus, Jun 23 '23 at 17:22

Red-eye removal in python using opencv

0 Answers0