This paper proposes video coding control methods which greatly improve image quality for videoconference or videophone systems. The proposed methods utilize frame and block classifications to incorporate subjective characteristics. Frame classification makes it possible to tune quantization step sizes according to which is more important regarding each processing frame, motion smoothness or image clarity. Block classification is used to adjust encoding parameters according to whether the block belongs to background regions or not. combining the two proposed method, quantization step sizes are controlled to achieve high image quality according to scene and block contents. Moreover, without unnecessary bit consumption, coding distortion is reduced in uncovered background regions, which appear after moving objects.
展开▼