A CS Image Denoising and Reconstruction Method Based on Hyperspectral Total Variation

Abstract

The present invention proposes a CS image denoising and reconstruction method based on hyperspectral total variation. The steps are as follows: initialize the reconstructed image, iterate the index value and the noisy observation value; use the noisy observation value to iteratively update the obtained reconstructed image to obtain Estimates; enter estimates separately into l 1 The intermediate reconstructed image is obtained from the CS reconstruction model of norm and HTV; the Starlet transform is used to sparsely represent the intermediate reconstructed image to obtain the Starlet coefficients; the new threshold operator and the improved BayeShrink threshold are used to de-noise the Starlet coefficients to obtain the curvelet coefficients; Perform Starlet inverse transformation on the curvelet coefficients to obtain the reconstructed image; determine whether the iteration stop condition is met, and loop iteration. The present invention can effectively protect feature information such as details and textures in the image while removing most of the noise information in the high-noise image, is simple to implement, has strong robustness, and effectively solves the problem of denoising and reconstruction of high-noise images. .

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a CS image denoising and reconstruction method based on hyperspectral total variation, which is used for denoising and reconstruction of high-noise images, and realizes high-efficiency high-resolution images under high-noise conditions. denoising capability. Background technique [0002] With the continuous progress of modern science and technology, CMOS / CCD sensor technology has also been developed rapidly, and the impact on us is mainly reflected in two aspects: (1) image quality is getting higher and higher; (2) image resolution Higher and higher. Although high-resolution images can bring us high visual enjoyment, they also bring new challenges to the field of image processing. When shooting high-resolution images at night, due to the influence of the night environment, the image data usually obtained contains a lot of noise information; in addition, when sh...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the technical problem of poor performance of denoising and reconstruction of high-noise images in existing CS reconstruction methods, the present invention proposes a CS image denoising and reconstruction method based on hyperspectral total variation, which uses the established CS reconstruction model to iteratively update the reconstructed image , use the Starlet transform to sparsely represent the high-noise image to obtain the Starlet coefficients, use the designed new threshold operator to filter the Starlet data of the obtained image in each iteration process, and effectively protect the details in the image while removing the noise Feature information can effectively realize high-quality reconstruction of high-resolution images under high-noise conditions

Images