Nonlocal self-similarity and sparse representation-based remote-sensing image denoising method

Abstract

The invention provides a nonlocal self-similarity and sparse representation-based remote-sensing image denoising method which comprises a dictionary learning process and an image reconstruction process, wherein the dictionary learning process comprises the step of constructing a group for each image block in an image; each group consists of nonlocal blocks with similar structures and adaptively learns a dictionary; the image reconstruction process comprises the step of calculating a spare coefficient of each group through an iteration shrinkage / thresholding algorithm, and obtaining a denoised image by using the dictionaries of the groups. According to the nonlocal self-similarity-based image denoising method, the nonlocal self-similarity of most images is used, and such structural self-similarity information is added into image denoising, so that a better denoising effect is achieved.

Description

technical field [0001] The invention relates to the technical fields of image processing and remote sensing, more specifically, to a method for denoising remote sensing images with sparse representation based on non-local self-similarity. Background technique [0002] With the rapid development of pattern recognition, image processing and computer vision technology, the development of remote sensing technology is changing with each passing day. However, remote sensing images are often disturbed by various noises during the acquisition and transmission process, which affects the image quality to a certain extent. , so that the sharpness of the image is significantly reduced, which has a great impact on the recognition of remote sensing images, target detection and segmentation. In order to improve the quality of the image and provide a basis for later image analysis, recognition, and higher-level processing, it is necessary to filter out the noise pollution in the image. Den...

AI Technical Summary

Problems solved by technology

Generally, the dictionary learning algorithm can be used to learn the dictionary. The K-SVD dictionary learning algorithm can adaptively learn a dictionary for noisy images according to the image, but it needs to train a global dictionary, and the time complexity is high.

[0004] The current denoising algorithm based on sparse representation of remote sensing images only considers the sparse representation of each image block itself, but does not take advantage of the possible geometric similarity between image blocks, which affects the denoising effect.

Images