Improved Temporal Scalable Video Coding Based on Low-Delay Hierarchical Dual Reference P-picture Prediction Structure

Temporal scalability is a very important part in scalable video coding (SVC). The traditional MCTF (motion compensated temporal filtering) algorithm, hierarchical B-picture, low-delay hierarchical P-picture have been adopted in H.264 SVC´s extension gradually. Our researching focus on low-delay hierarchical P-picture structure, the traditional low-delay hierarchical P-picture structure have some shortcoming: if the GOP (grope of picture)´s frame number is increasing, there are more P frame need to be stored in memory for later P frame´s reference, which would increase the complexity of decoding process and increase the delay of coding. In this paper, an improved hierarchical dual reference P-picture structure have been proposed to increase the encoding efficiency even lower the coding delay and memory requirements. Experimental results show that this method increasing the average PSNR value about 0.05 db, reducing the video bit-rate of 2% at the same time without increases any P frame storage.