Smooth quality streaming with bit-plane labeling

摘要

Bit-plane coding techniques have been proposed as an efficient way to achieve the goal of fine granularity scalability (FGS). Both in image and video coding standards such as JPEG 2000 and MPEG-4, bit-plane coding techniques have been utilized to code a universal bit-stream for various bandwidth applications. With the bit-plane coding, a bit-stream can be truncated according to the available bandwidth. Hence, the video quality can be very sensitive to the network bandwidth variation especially while streaming a video over the Internet or a wireless link. For overall human perception, a consistent and smooth quality video is much better than to have high video quality in just a short period but with highly quality variation in general. We classify the quality variation into two categories: inter-frame and intra-frame quality variation. The inter-frame quality variation may cause by the constant bit rate coded (CBR) base layer video or the network bandwidth variation, while the intra-frame quality variation causes mainly by the various motion types. For example, moving objects are usually more difficult to be encoded. This paper focuses on the inter-frame quality variation because quality variation is more significant between frames than within a frame. An enhancement layer labeling algorithm is proposed to reduce the inter-frame quality variation, and a dynamic bit-plane truncation scheme is utilized to maintain the smooth streaming video quality. The proposed algorithm has no overhead in the decoder, and the experimental results portray that our proposed algorithm can greatly eliminate the inter-frame quality variation.