This is the Jupyter notebook for automatic watermark removal code. We follow the scripts till watermark estimation and then try to solve the watermarking problem.
Here comes the tougher part. We have an initial estimate for the watermark
Wm, cropped_gx, cropped_gy and
est. Using this, we solve the problem of Multi-image matting and reconstruction. We have a list of images,
images_cropped, and we have to estimate
alpha by reiterating the following 3 steps:
Finally, we do a Matte and blend factor initialization to determine the constant and normalized matte.