Abstract: An improved reconstruction method to enhance the image resolution for multi-frame video sequence is proposed based on jointly making use of the techniques of Maximum A Posteriori Probability (MAP) and Project Onto Convex Sets (POCS). At first, a simplified spatial-domain estimator is employed by introducing the preconditioned conjugate gradient method to forecast the original high-resolution image. Then, under the wavelet-domain, two different convex sets and projection operators are designed from the aspects of inter-frame and intra-frame to extract the details hidden among the adjacent observed low-resolution frames. At last, taking advantage of the relationship between the neighborhood factors in the spatial estimator to put constraints on the potential solutions of the POCS, it can get uniqueness optimal solution quickly. The results of the simulation experiment and practical application of supervisory system demonstrate that the proposed method not only outperforms the other methods on the aspects of PSNR (Peak Signal-to-Noise Ratio) and visual quality of the reconstructed sequence, but also has the advantages of rapid convergence and great robust.

Key words: computer application, super-resolution reconstruction, preconditioned conjugate gradient (PCG), project onto convex sets (POCS), video sequence