Immune-clonal-selection-based nonsubsampled contourlet domain image denoising method

Abstract

The invention relates to the application of an image processing technology in the field of image denoising, in particular to an immune-clonal-selection-based nonsubsampled contourlet domain image denoising method, which is characterized by comprising the following steps of: (1) inputting a noisy image X, and performing L-layer nonsubsampled contourlet domain decomposition on the noisy image X to obtain high-frequency directional sub-bands {Dl,i(m,n),} in different scales and a low-frequency sub-band AL(m,n), wherein l is more than or equal to 1 and less than or equal to L-1, i is more than or equal to 1 and less than or equal to kl, kl is the number of the high-frequency direction sub-bands in the scale of 2-l, Dl,i(m,n) represents a coefficient at a pixel position (m,n) on the ith high-frequency directional sub-band of the noisy image in the scale of 2-l, and L is 3 to 5; and (2) searching for an optimal denoising threshold value {Tl,i} of each high-frequency directional sub-band in the different scales by using an immune clonal selection method, wherein l is more than or equal to 0 and less than or equal to L-1, and i is more than or equal to 1 and less than or equal to kl. Compared with the prior art, the invention searches for denoising threshold values by adopting the immune clonal selection method without knowing the exact characteristics of image noise, so that better denoising threshold values can be found; and due to the adoption of nonsubsampled contourlet transform, jitter distortion caused by the deficiency of translation invariance of a transform tool can be effectively avoided.

Description

technical field [0001] The invention relates to the application of image processing technology in the field of image denoising, in particular to a non-subsampling contourlet domain image denoising method based on immune clone selection. Background technique [0002] Images are usually disturbed by noise during the process of acquisition, compression and transmission. The existence of noise is not conducive to people's processing and interpretation of images, so before analyzing image data, it needs to be denoised. Typical image denoising methods include methods based on spatial domain and methods based on transform domain. The methods based on space domain include mean filter, median filter, Wiener filter and so on. These spatial domain-based techniques tend to blur the details of the image while filtering out noise, and the filtering performance is largely dependent on the size of the selected filtering window. The method based on transform domain utilizes the sparsity o...

AI Technical Summary

Problems solved by technology

Although the wavelet transform can optimally approximate the point-like singularity, it is not optimal in high-dimensional situations, or the "sparse" image representation method, which cannot effectively capture the edges, contours, curves, etc. in the image. Dimensional information, affecting image denoising effect

Images