Psychovisual video coding using wavelet transform

Visual communications services are now making a significant impact on modern society. Video conferencing, HDTV and multimedia are just examples where this technology is being used to good effect. Communicating using video signals does, however, require a large volume of data to be transmitted, and even with modern high-bandwidth communication links this can be expensive. This requires the implementation of efficient video coding and compression schemes. This thesis investigates both image and video coding compression schemes and aims to develop a scheme with the highest possible performance. In image coding there are two main types of compression: statistical and psychovisual. This thesis concentrates on the latter, since it is shown that psychovisual techniques, in general, provide greater levels of compression than statistically based methods. The standardised technique for video coding uses psychovisual compression of the coefficients of the discrete cosine transform (DCT). Despite being an international standard for low bit rate video coding the DCT suffers from a number of drawbacks. Firstly, the psychophysical and psychological models of the human visual system (HVS) are based on a multiresolution approach whereas the basis functions of the DCT are fixed in resolution. Secondly the basis functions of the DCT only possess good localisation properties in the frequency domain and not the spatial domain, a characteristic that blurs edges and discontinuities in an image. By contrast the wavelet transform is a multiresolution approach and its basis functions can possess good localisation properties in both the spatial and frequency domains. Furthermore, due to the excellent localisation properties of the wavelet function most of the transform coefficients are practically zero and the use of wavelet transform can be expected to achieve a higher compression ratio than the DCT. This thesis therefore investigates psychovisual transform coding using the wavelet transform instead of the DCT.