Locally connected Hopfield multilayer network for associative memories storing 256 levels grey-scale images

In this paper a new design procedure to design associative memories storing grey-scale images is presented. It is based on a locally connected multilayer Hopfield neural network, with both intra-layer and inter-layer connections. The proposed architecture is an evolution of a previous work based on the decomposition of the image with 2L gray levels into L binary patterns, stored in L uncoupled neural networks: that allows to store images with the commonly used number of 256 gray levels. The approach presented in this paper is based on a suitable decomposition of the grey-scale image into binary images, stored in Hopfield neural networks with bipolar states. The learning algorithm, used to store the binary images, guarantees asymptotic stability of the stored patterns, has a low computational cost, and allows to control the precision of the connection weights. Some design examples are presented to show the effectiveness of the proposed method.