Foreground/background video coding scheme

Abstract

This paper presents the use of segmentation to improve the subjective quality of the sequence produced by very low bit rate video coding system. Through this approach, each frame of the source sequence is first segmented into two non-overlapping regions, namely foreground and background. These two regions are then encoded using the same coder but with different quantization step-sizes. In this way, the image quality of the foreground region can be improved at the expense of encoding the unimportant background region at lower quality. Currently, our work focuses on integrating this approach into the H.263 coder, primarily for the videotelephony application. In this paper, we describe the working implementation, and also demonstrate the improved subjective quality of the coded sequence achieved.