CLAHE Implementation produces unexpected output

Question

I am trying to implement CLAHE(contrast-limited adaptive histogram equalization) algorithm using python3, numpy and cv2. I am using the slower non-interpolation method. Just dividing image to tiles and trying to equalize histograms of sub-images to enhance contrast. But the output is is very weird. I looked up some github repo's and find one which my histogram and clipped_histogram_equaliziation method is same with that methods from the repo.

Repo link: https://github.com/asalmada/x-ray-images-enhancement/blob/master/src/algorithms/clahe.py

Below you can see my code, cv2 output and my output. I spent hours and I can't see what I am missing

def clahe(img, tile_size: int = 8, clip_limit: int = 40):
    width = img.shape[1]
    height = img.shape[0]

    padding_size = tile_size // 2
    padded_img = padding(img, padding_size)
    p_width, p_height = padded_img.shape[1], padded_img.shape[0]
    clahed_img = np.zeros((p_height, p_width), 'uint8')


    for i in range(padding_size, p_height - padding_size):
        for j in range(padding_size, p_width - padding_size):
            tile = padded_img[i-padding_size:i +
                padding_size, j-padding_size:j+padding_size]
            hist = clipped_histogram_equalization(tile, 255, clip_limit=40)
            clahed_img[i][j] = hist[padded_img[i][j]]

    enhanced_img = clahed_img[padding_size:p_height -
                              padding_size, padding_size:p_width - padding_size].astype(np.uint8)

    return enhanced_img



def histogram(data):
    pixels, count = np.unique(data, return_counts=True)
    hist = col.OrderedDict()

    for i in range(len(pixels)):
        hist[pixels[i]] = count[i]


    return np.array(list(hist.values())), np.array(list(hist.keys()))


def clipped_histogram_equalization(region, max_val, clip_limit):

    hist, bins = histogram(region)
    n_bins = len(bins)

    # Removing values above clip_limit
    
    excess = 0
    for i in range(n_bins):
        if hist[i] > clip_limit:
            excess += hist[i] - clip_limit
            hist[i] = clip_limit

    ## Redistributing exceding values ##
    # Calculating the values to be put on all bins
    for_each_bin = excess // n_bins
    # Calculating the values left
    leftover = excess % n_bins

    hist += for_each_bin
    for i in range(leftover):
        hist[i] += 1
    
    # Calculating probability for each pixel
    pixel_probability = hist / hist.sum()
    # Calculating the CDF (Cumulative Distribution Function)
    cdf = np.cumsum(pixel_probability)
    cdf_normalized = cdf * 255

    hist_eq = {}
    for i in range(len(cdf)):
        hist_eq[bins[i]] = round(cdf_normalized[i])

    return hist_eq

And here is my main method:

img = cv.imread("test6.jpg", 0)
enhanced_img = clahe(img)
cv.imshow("Image after CLAHE", enhanced_img)

Finally, the outputs: