by Ayman Darwish
Supervisor: Dr Tim Ferguson
A scalable video codec based on the discrete wavelet transform is developed and implemented from scratch. The codec produces an embedded bitstream for each frame of the video sequence by means of the EZW algorithm. This bitstream allows for flexible quality-scalable decoding, in which a decoder can stop decoding the bitstream as soon as the desired quality level has been reached. The progressive quality improvement with further decoding is an inherent property of the embedded bitstream that comes at the cost of no overhead. An overlapped block motion compensation system is employed to overcome issues associated with applying the discrete wavelet transform to prediction error frames. Results are generated for each component of the codec, examining sub-pel accuracy, OBME algorithms, and fast block-matching searches.
Download the thesis in PDF format: thesis.zip 900KB