9

I have the images of cartoons, that were compressed. The example:

Original image

They have such noise, that is not easy to delete.

enter image description here

Even though the pixels are on the gray background, the noise pixels can be of very different colors and if I look at the histogram (of grayscale image) of such image I see a lot of bins around the bin of main color. But I can't just delete the colors of value, that are not of main color, because there is a risk to delete some important features (as eye, for example). Also I've tried to posterize image (example presents 8 colors), but some pixels still remain.

enter image description here

Also I've tried median filter, but it is not able to vanish such heavy noise ( I use 3*3 filter).

Please, give some methods that can effectively remove the noise in this case. I would be grateful for any help!

Ann Orlova
  • 191
  • 4
  • 1
    There are numerous papers specifically on denoising cartoons. There are also practical guides on sites like doom9 that cover the practical aspects. – Emre Mar 04 '13 at 17:06

1 Answers1

1

The Total Variation (TV, a.k.a. Rudin-Osher-Fatemi functional) denoising wold help here. You can apply it on every color layer or look for a vectorized version if th eresults ar enot good enough for you.

Good references of tsable and easy to implememnt algorithms for TV denoising are for example:

Both references have good performance and are easy to implement from the papers.

sansuiso
  • 3,937
  • 18
  • 27