A Method of 0 l Restriction with Multi-scale Product for Image Denoising

Abstract

In order to eliminate image noise and retain more image feature information, this paper introduced an image denoising method based on sparse representation and multidimensional calculation under 0 l restriction, using 0 l restriction to optimize dictionary atoms according to the randomness of image noise and directly eliminate atoms that have been used less frequently to improve calculation effectiveness. Besides, since residual images may also contain useful information, noise reduction was conducted against residual images by applying wavelet transform modulus maximum and multi-scale product calculation, the result of noise reduction realized through dictionary sparse representationwas integrated with the result of denoising residual images firstly to obtain denoised images. The result of the experiment shows that comparing with K-SVD algorithm, the time complexity of the algorithm used in the paper is lower and therefore can obtain higher PSNR value.