In this paper, based on projection onto convent sets (POCS) principle, expatiate the elements and the approach of the projection onto convent sets principle. A modified method of POCS has been put forward, for raise the quality of the reconstructed image. It will be the super-resolution reconstruction iterative elementary value that the bilinear interpolation image of be deal with the arithmetic of the iterative blind deconvolution in frequency. Because of no elementary image, we just have average grads and information entropy the results to be compared. Such the experiment result shows that this algorithm which is in this paper could significantly improve the quality of the reconstructed image compare with traditional POCS algorithm.