Fully scalable subband/wavelet coding

Due to the growth and success of the Internet and wireless communication, as well as increasing demand for multimedia services, streaming media over the Internet and mobile/wireless channels has drawn tremendous attention from both academia and industry. The MPEG-21 Part 7: Digital Item Adaptation (DIA) is being standardized to achieve the key goal of multimedia delivery: To be created and shared, always with the agreed/contracted quality, reliability, and flexibility. Scalable coding is a both effective and efficient solution for this purpose. By embedding lower resolution or quality bitstreams into higher resolution or quality bitstreams, scalable coding is a new approach to represent and prioritize data. DIA can be easily realized by selectively transmitting sub-bitstreams.; In this thesis, a fully scalable video coding algorithm with combined temporal, spatial and quality scalability is designed. Since subband/wavelet coding inherently provides multi-resolution representation, a motion compensated 3D (spatiotemporal) subband/wavelet transform is utilized to realize temporal and spatial scalability. Motion compensated temporal filtering (MCTF) plays an essential role in this 3D transform. It influences coding efficiency and also temporal scalability, since temporal low frequency subbands will be the embedded lower frame rate videos. I introduce sub-pixel accurate MCTF with lifting scheme. The lifting scheme makes it easy to introduce sub-pixel accuracy into MCTF, while retaining perfect reconstruction. We observed significant improvement up to 1/8 pixel accuracy at CIF resolution. I also design a pixel-based detection algorithm to locate the real to-be-covered and uncovered pixels. Poorly connected blocks are detected and converted to I blocks to protect the lower frame rate videos free from motion artifacts. Bidirectional MCTF is also realized to increase coding efficiency. An entropy coder EZBC is incorporated to encode generated subband coefficients.; Based on these techniques, I develop a complete video coding system MC-EZBC, matching the requirements of MPEG-21 DIA. MC-EZBC includes three parts: Pre-encoder, transcoder/extractor and decoder. In the pre-encoder, a 3-D transform naturally partitions video data into a range of spatiotemporal resolutions and precision levels. The transcoder/extractor truncates bitstreams based on the user preferences and network properties. At the receiver, the decoder decodes the received bitstream. Such a coding system provides high coding efficiency, and flexibility and computational simplicity in transcoding.