Integration of histograms in the NL-Mean algorithm for image denoising.


Revision as of 13:08, 6 June 2018 by Bot (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)


Pictures have noise and to reduce it, some algorithms are applied. One of them is the NL-Mean algorithm, implemented as a patch-based strategy. The main idea is to denoise a local region based on other similar neighboring patches. A spatial metric (the Euclidian distance) is used to determine the similarity between patches. The main objective of this project is to incorporate the information related to the statistical distribution of colors within a patch, thanks to the use of histograms, to potentially highlight characteristics that have been missed with the standard strategy. It may give some interesting information which could be helpful to improve picture denoising. This method adds robustness regarding any transformation which does not affect the shape of histograms (overall change of luminance or rotation) and can allow the use of patches which could have been discarded by the standard strategyafter proper post-processing of the patch.