Region of interest based H.263 compatible codec and its rate control for low bit rate video conferencing

This paper presents a region of interest (ROI) based H.263 compatible video codec, which combines the idea of object-based coding from MPEG4 visual into the traditional block-based H.263 codec. A face detection and tracking scheme with very low complexity is proposed to segment human face from video conferencing sequences in real-time. With the segmentation information, the ROI based codec and its associated rate control schemes are designed. For VBR video, the proposed rate control is a joint frame layer and macroblock layer scheme. For CBR video, a macroblock layer rate control is proposed. TMN8 is adopted as the platform, and the modified quantization mode in Annex T of H.263 is adopted to achieve flexibility in assigning quantization parameters among different macroblocks.