A spectral algorithm for removing salt and pepper from images
Olivier Rioul · 2002
This paper presents an elegant solution to the impulse noise cancellation problem for images. The approach taken here is based on the theory of error-correcting (BCH) codes. Impulses are considered as "errors". The process is in two steps: First, find the error locations in the image; then, correct the error values. A first pass is made on the lines of the image, a second pass is then performed on the columns. Next the same procedure is repeated. In most cases, impulse noise is completely removed after a few iterations are performed.