Image compression using multi-layer neural networks

A new neural network data compression is presented. This work extends the use of 2-layers neural networks to multi-layer network. Results show the performance superiority of multi-layer neural networks compared to that of the 2-layer one especially at high compression ratios. To overcome the long training time required for multi-layered network, a recently-developed training algorithm has been used. A modified feed-back error is proposed to further reduce the training time and to enhance the image quality. Also, a redistribution of the gray levels in the training phase is proposed to make the minimization of the mean square error more related to the human vision system.