Image restoration method based on mixed sparse prior model

The invention discloses a natural image restoration method, and mainly aims to solve the problem that the objective restoration effect and the subjective visual effect cannot be satisfied in the priorart. The implementation scheme comprises the following steps: 1) initializing a restored image and the number of cycles; 2) constructing and training a 12-layer convolution neural network; 3) calculating the initial estimated value of the initial restored image; 4) calculating the initial estimated value of a sparse feature map from the initial restored image; 5) blocking the initial estimated value of the initial restored image, and calculating the weights of blocks; 6) calculating the non-local estimated value of the sparse feature map from the weights of the blocks; 7) calculate the prioriestimated value of the sparse feature map from the initial estimated value and the non-local estimated value of the sparse feature map; 8) constructing an objective function for solving the restoredimage based on the priori e