Abstract
In recent years, the need of transmitting multimedia contents over TCP/IP on the Internet is growing rapidly. The conventional video compression techniques are concentrated on optimizing the coding performance at a given bitrate. However, these video coding techniques are not suitable for multimedia streaming over the Internet due to the bandwidth differences of heterogeneous networks and the bandwidth variations with time. In order to transmit multimedia contents over the Internet, the bitstream should be partially decodable at any bitrate to reconstruct a video signal with the optimized quality at that bitrate. This is the objective of fine granularity scalability (FGS) video coding techniques in the Amendment of MPEG-4.The bitplane coding of the DCT residues was adopted by the MPEG committees. In this thesis, we propose an embedded-quantizer based encoder to achieve fine granularity scalability. The concept of embedded-quantizer is to quantize the DCT coefficients once, and then separate the quantized coefficients into several parts and code each separated part individually. If all separated parts can be received and decoded, the performance will be the same as if there were no separation.The performance of the embedded-quantizer based encoder is evaluated. We proved that this structure could achieve fine granularity scalability although the performance is merely approaching the structure provided by FGS reference software. We believe that its performance could be better if the bitstream syntax was optimized.