Study of wavelet decompositions for image/video compression by software codecs

Although other workers have examined the quality/compression tradeoff for various wavelet approaches, as far as we are aware, no adequate study has been made of the tradeoff between fidelity, processing requirement, and bandwidth. This is crucial to the design of software-only codecs dependent on utilising the main system processing resource. This paper describes on-going work aimed at defining an optimum architecture for such components, and presents results of a systematic study of the standard and non-standard decomposition approaches in conjunction with the Haar wavelet transformation For all images the non-standard decomposition is markedly superior to the standard decomposition achieving over four times the compression for similar quality of reproduction. The algorithm for the non-standard decomposition and the standard decomposition are both of outstandingly low computational complexity with the non-standard being slightly quicker. The results show that the non-standard decomposition in conjunction with the Haar wavelet is an excellent choice as the algorithmic base for a software-only codec.