The neural network modelled POCS method for removing blocking effect

This paper proposes a new method for real-time realization on the blocking effect elimination. This is achieved by training a feed-forward single-layer neural network (FFSLN) to restore block boundaries of JPEG encoded images. The reconstructed image of the iterative projection onto convex sets (POCS) method instead of the original image is chosen as the target output in this proposed method. Computer simulation result demonstrates the superiority of the new method as compared with the original POCS iterative recovery method