Sparse coding method capable of solving problem of similarity retention in super-resolution reconstruction of image

The invention relates to a sparse coding method capable of solving the problem of similarity retention in super-resolution reconstruction of an image. The sparse coding method comprises the following steps of: (1) a training stage: randomly extracting high-resolution image blocks and low-resolution image blocks, and carrying out preprocessing; and training a joint dictionary by using a Laplace sparse coding method so as to obtain a high-resolution dictionary and a low-resolution dictionary; and (2) a testing stage: reading a test image set, loading the high-resolution dictionary and the low-resolution dictionary, processing test image blocks and reconstructing high-resolution image blocks; finding an image which is the most similar with a high-resolution image by using a gradient descent method; and outputting a high-resolution image. According to the method disclosed by the invention, instability of sparse coding processing can be reduced, so that a better super-resolution reconstruction effect is achieved.