Image denoising using overcomplete DCT Dictionary



This function denoises an image by sparsely representing each block with the overcomplete DCT Dictionary, and averaging the represented parts. 

Detailed description can be found in “Image Denoising Via Sparse and Redundant representations over Learned Dictionaries”, (appeared in the  IEEE Trans. on Image Processing, Vol. 15, no. 12, December 2006).


INPUT ARGUMENTS : Image - the noisy image (gray-level scale)

                   sigma - the s.d. of the noise (assume to be white Gaussian).

                   K - the number of atoms in the representing dictionary.

Optional argumeters:             

                  'blockSize' - the size of the blocks the algorithm

                       works. All blocks are squares, therefore the given

                       parameter should be one number (width or height).

                       Default value: 8.

                  'errorFactor' - a factor that multiplies sigma in order

                       to set the allowed representation error. In the

                       experiments presented in the paper, it was set to 1.15

                       (which is also the default value here).

                  'maxBlocksToConsider' - maximal number of blocks that

                       can be processed. This number is dependent on the memory

                       capabilities of the machine, and performances’

                       considerations. If the number of available blocks in the

                       image is larger than 'maxBlocksToConsider', the sliding

                       distance between the blocks increases. The default value

                       is: 250000.

                  'slidingFactor' - the sliding distance between processed

                       blocks. Default value is 1. However, if the image is

                       large, this number increases automatically (because of

                       memory requirements). Larger values result faster

                       performances (because of fewer processed blocks).

                  'waitBarOn' - can be set to either 1 or 0. If

                       waitBarOn==1 a waitbar, presenting the progress of the

                       algorithm will be displayed.

OUTPUT ARGUMENTS : IOut - a 2-dimensional array in the same size of the

                   input image, that contains the cleaned image.

                    output - a struct that contains that following field:

                       D - the dictionary used for denoising


Image Denoising via Dictionary Learning and Structural Clustering



There are no reviews yet.

Be the first to review “Image denoising using overcomplete DCT Dictionary”

Your email address will not be published. Required fields are marked *