Lossy image compression using a modular structured neural network
E. Watanabe, Katsumi Mori · 2002
A lossy compression method for gray images is proposed on the basis of a modular structured neural network. This modular structured neural network consists of multiple neural networks with different block sizes (the numbers of input units) for the region segmentation. By the region segmentation procedure, each neural network is assigned to each region such as the edge or the flat region. From simulation results it is shown that the proposed compression method gives better compression performance compared with the conventional compression method using a single neural network.