Video coding with adaptive vector quantization and rate distortion optimization

The object of this dissertation is to investigate <br>rate-distortion optimization and to evaluate the prospects of adaptive <br>vector quantization for digital video compression. <br>Rate-distortion optimization aims to improve compression performance <br>using discrete optimization algorithms. We first describe and classify <br>algorithms that have been developed in the literature to date. <br>One algorithms is extended in order to make it generally applicable; the correctness of this new procedure is proven. Moreover, we compare the complexity of the aforesaid algorithms, first <br>implementation-independent and then by run-time experiments. <br>Finally, we propose a technique to speed up one of the aforementioned algorithms. <br>Adaptive vector quantization enables adaption to sources with unknown or non-stationary statistics. This feature is important for digital <br>video data since the statistics of two subsequent <br>frames is usually similar, but in the long run the general statistics of <br>frames may change even if scene changes are neglected. <br>We examine combinations of adaptive vector quantization with various state-of-the-art video compression techniques. <br>First we present an adaptive vector quantization based codec that is able to encode and decode in real-time using current PC technology. This codec is rate-distortion optimized and adaptive vector quantization is applied in <br>the wavelet transform domain. The organization of the wavelet coefficients is then made more efficient using adaptive partition techniques. Moreover, the main adaptability mechanism of adaptive vector quantization, the so-called codebook update, is studied. <br>Finally, a combination of adaptive vector quantization and motion compensation is taken into consideration. We show that for very low bitrates adaptive vector quantization performs on prediction <br>residual frames better or at least as well as discrete cosine transform coding.

Author dc:creator

• Wagner, Marcel

Contributors dc:contributor

• Saupe, Dietmar

• Videocodierung
• Vector quantization
• Rate/Distortion optimization
• video coding

Repository record source_url

https://freidok.uni-freiburg.de/data/209

OAI identifier oai:identifier

oai:freidok.uni-freiburg.de:209