MPEG-4 FGS(Fine Granularity Scalability) coding is a scalable coding with fine granularity in bitrate so that it can be adaptively used in channels with changing bitrate as in Internet. It has an optional selective enhancement(SE) algorithm to enhance the subjective quality in the limited bitrates. The rectangular region-based SE(RSE) is an advanced SE that has the advantage of reduced bit-increase in selective enhancement.
In this thesis, the RSE-FGS video coding algorithm is implemented by software in both encoding and decoding, including the determination of the rectangular region of interest(ROI). It is based on the frame difference and macro-block of image occurred in the base layer coding. To further reduce the coded bits of the enhanced region at the mid-bit-plane layers, a new concept of partial enhancement(PE) of bit-planes is introduced.
It is shown by simulation that the proposed PE algorithm provides a significant gain over the simple RSE-FGS algorithm in the middle range of bitrates.
