Removing watermark using opencv in python

Question

I have used OpenCV and Python to remove a watermark from an image using code below.

import cv2
import numpy

src = cv2.imread('src.jpg')
mask = cv2.imread('mask.jpg')
save = numpy.zeros(src.shape, numpy.uint8) 

for row in range(src.shape[0]):
    for col in range(src.shape[1]):
        for channel in range(src.shape[2]):
            if mask[row, col, channel] == 0:
                val = 0
            else:
                reverse_val = 255 - src[row, col, channel]
                val = 255 - reverse_val * 256 / mask[row, col, channel]
                if val < 0: val = 0

            save[row, col, channel] = val

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

Here are the src, mask files and what I get from the code

I tried to neutralize the original image watermark with a white background inverse watermark image.

But now it makes no progress and I don’t what happened to it.

I googled a bit and found some info about it, but in my case I have a mask.
How can I achieve it with my current code? Any help is appreciated.

source

mask

result

Asked By: wwwww135781012

||

Source

Answer 1

ok, I didn’t understand your algorithm of "neutralizing":

reverse_val = 255 - src[row, col, channel]
val = 255 - reverse_val * 256 / mask[row, col, channel]

But I can tell you, why it doesn’t work. When you read the mask image mask = cv2.imread('mask.jpg') you read it as usual image with white background and the watermark. You should make an binary image from this, so you will undestand which pixels you need to neutralize:

threshed = cv2.inRange(mask, 0, 254)

And you will get something like this:

So now you need to neutralize the "white" pixels of mask on the original image.
(BUT! You should remember that binary image and grayscale image have only one channel).
As I said before: I didn’t understand your algorithm of "neutralizing" because when I use it – I get an image like this:

So I recommend you just to make all the pixels white:

And the full code:

import cv2
import numpy


src = cv2.imread("src.jpg")
mask = cv2.imread("mask.jpg", cv2.IMREAD_GRAYSCALE)
save = numpy.zeros(src.shape, numpy.uint8)

threshed = cv2.inRange(mask, 0, 254)


def get_reversed(values: tuple) -> tuple:
    return 255 - values[0], 255 - values[1], 255 - values[2]


def get_processed(values: tuple, mask: int) -> tuple:
    return 255 - values[0] * 256 / mask, 255 - values[1] * 256 / mask, 255 - values[2] * 256 / mask


for row in range(src.shape[0]):
    for col in range(src.shape[1]):
        if threshed[row, col] != 0:
            # save[row, col] = get_reversed(src[row, col])
            save[row, col] = (255, 255, 255)
            # reverse_val = get_reversed(src[row, col])
            # val = get_processed(reverse_val, mask[row, col])
        else:
            save[row, col] = src[row, col]


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

Answered By: crackanddie

Removing watermark using opencv in python

Question:

Answers: