An Object of Interest based Segmentation Approach for Selective Compression of Video Frames

Abstract

The automatic segmentation of objects of interest is a new research area with applications in various fields. In this paper, the object segmentation method is used for content based video management and compression of video frames for video conferencing. The face region, which is the object of interest in the video frames, is identified first using a skin color based algorithm. The face regions are then extracted and encoded without loss, while the nonface regions and the non-face frames are quantized before encoding. Results show that the decompressed video has an improved quality with the proposed approach at low bit rates.