Algorithm And Implementation Of H.264/SVC Multiplexing

Currently in the background of the integration of telecommunication networks,cable TV networks and Internet, the user terminal becomes diverse and the networktransmission capacity changes as well. The traditional way of encoding can onlygenerate one video encoding consequence, which means fixed frame rate in timedomain, fixed spatial resolution in spatial domain, and it is unable to adapt to a varietyof application environments. If transmitting the video in different networkenvironments is demanded, the only way to realize it is through the transcodingtechniques, which means decoding and encoding the video. SVC is one of theeffective solutions to this problem. SVC can produce video that enables differentdomain scalable, temporal scalable, spatial scalable, and quality scalable.Although current coding technologies have researched deeply in videocompression efficiency, it also brings the problem of huge bit-rate fluctuations.However network transmission design is often based on the peak value of the videostream, so the benefit brought by compression is actually wasted. Hence study of SVCmultiplexing to improve bandwidth utilization is very necessary. The experiment wasconducted for SVC multiplexing of the MPEG-2TS as well as T-DMB, and theresults show that the multiplexing algorithm does not result in any overflow orunderflow, and effectively improves the bandwidth utilization.In addition, the article also optimized the original selection method of SVCquantization parameter in the process of SVC quantization.The algorithm takes notonly the effect coding structure, but also the effect the video content, andcompression result into consideration. Depends on whether scene change occurred ornot, reasonable quantization parameter adjustments should be taken. Therefore it cangreatly improve the coding performance by providing high quality reference picture.The experiment also tested a large number of sequences. In those sequences, similarresults are obtained using the proposed method. Compared with JSVM quantizationmethod, this method can achieve a PSNR gain of0.1~0.25dB.