Color Image Denoising via Sparse 3D Collaborative Filtering With Grouping Constraint In Luminance-Chrominance Space
K.Dabov,A.Foi, V.Katkovnik and K.Egiazarian, International conference on Image Processing, ICIP,vol 1, Pg 313-316, Oct 2007
In this paper , color image denoising is based on enhanced sparse representation in transform domain. This enhancement is achieved by grouping similar 2D fragments(blocks) into 3D data arrays or groups. This is done in the luminance-chrominance channel.The RGB image corrupted by additive white gaussian noise is transformed into luminance-chrominance color space. Grouping is done by block matching, by choosing the similar blocks by measuring the dissimilarity with the reference one. The dissimilarity smaller than the threshold is considered to be similar.Grouping similar 2D blocks into 3D array gives a high sparse representation of the signal in the transform domain.Shrinkage of the transform spectra results in noise attenuation and the inverse transform produces estimate of all the groups in the block.
This denoising algorithm shows good preservation of uniform areas, smooth intensity transitions, textures, repeating patterns,sharp edges and singularities. This method achieves good denoising performance in terms of both Peak signal-to-noise ratio(PSNR).
This denoising algorithm shows good preservation of uniform areas, smooth intensity transitions, textures, repeating patterns,sharp edges and singularities. This method achieves good denoising performance in terms of both Peak signal-to-noise ratio(PSNR).