Image Encoding by Multi-Spectral Associative Memory Neural Network

This work presents a new solution to overcome the obstacle of using Hopfield Neural Network with high level color images than binary images. This becomes a significant challenge in many applications with multi-spectral image recognition. Therefore, one may suggest adapting Hopfield model to perform high level image recognition by encoding the input data. The encoding will perform form all pixels in an image such that the rules of Hopfield model still valid and the output results should be agreed with converging. The important preliminary results of this work are represented by the advantages of using the new associative memory which based on Hopfield model for encoding high color images with any depth of pixels. In the encoding stage the number of colors can determine the proper level of the new technique. This leads to the ability of using the technique for compressing data easily along with the recognition operation.