New transforms and neural networks for image coding

Abstract

Image coding is the key technique for modern digital image communications. This thesis presents the development and analyses of new transforms and neural net-works for image coding. The first part of the thesis is devoted to the development and application of new transforms. These new transforms have been developed by the author to have at least one of the following features: lower implementation complexity or higher performance compared to the conventional discrete cosine transform when they are used for image data compression. The performance of the new transforms are assessed in terms of subjective as well as objective measures.