Image recovery method and system based on multi-scale random proximity algorithm

The invention discloses an image restoration method and system based on a multi-scale random proximity algorithm. The method comprises the following steps: inputting a Gaussian blur image with noise;constructing a tight frame filter induced by discrete cosine transform, and carrying out convolution on the Gaussian blurred image to obtain a tight frame wavelet coefficient of the Gaussian blurred image; optimizing the tight frame wavelet coefficient by using a balance sparse model, and obtaining an immobile point equation about the tight frame wavelet coefficient through an adjacent operator and a sub-differential method; and solving the fixed point equation to obtain an optimized tight frame wavelet coefficient, and obtaining a restored image after the transposition effect of a tight frametransformation matrix. According to the method, a random proximity algorithm based on a multi-scale structure is adopted, the technical problem of low calculation efficiency of large-scale image recovery is solved, and the ca