IMAGE COMPRESSION USING NEURAL NETWORKS

Work is described in which neural network techniques have been used to compress single and moving images. For the still images, two different types of neural network have been examined: the multi-layer perceptron and the Kohonen self-organizing network. In many of the applications in which there is a need for image compression, moving images are involved. It is shown how the Kohonen network can be used to compress moving images by utilizing the redundancy between consecutive frames. Also addressed is how neural networks may be made adaptive, taking into account changes in the scenes they are required to compress.