Automatic watermark removal


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.

Loading output library...
Loading output library...
Loading output library...

Image reconstruction


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 W and alpha by reiterating the following 3 steps:

  • Image-Watermark detection
  • Watermark update
  • Matte update

Finally, we do a Matte and blend factor initialization to determine the constant and normalized matte.