Potential function NN algorithm for image decomposition and compression

Abstract

This paper presents a neural network (NN) algorithm for image decomposition and compression based on potential functions. Our idea is to approximate the image block with orthogonal functions and respective potential functions whose coefficients will be weights in neural networks. The process is iterative and is implemented on the so-called Answer-in-Weights NN. We applied the compression method to various gray scale images and show its efficiency in comparison with JPEG. We conclude with a summary of future challenges and research directions.