基于MCTF的时域可分级视频编码技术

Temporal Scalable Video Coding Based on MCTF

现代可分级视频编码的时域可分级性是采用运动补偿时域滤波技术(Motion-Compensated Temporal Filtering,MCTF)来实现的,MCTF将若干连续的视频帧组成一个图像组进行小波分解以提供时域可分级性。由于不同视频序列的时域特征变化,传统的图像组大小固定的时域分解影响编码效率。本文提出了一种图像组大小自适应调整的时域小波分解算法,利用帧间预测效果来决定视频序列的时域特性,根据时域特性用于小波分解的图像组大小做自适应的调整。实验结果表明,新算法的编码性能比图像组大小固定的结构大大提高。

Temporal scalability of scalable video coding （SVC） is obtained by executing motion-compensated temporal filtering （MCTF）. MCTF decomposes a group of pictures（GOP） consisted of several successive frames to rapport temporal scalability by using wavelet transform. The coding efficiency might be reduced owing to the variations of the image characteristics of the various sequences in traditional fixed-size GOP. An adaptive-size GOP is proposed in this dissertation. This algorithm is to estimate the sequence＇s temporal characters by the aid of inter-prediction, through which the adaptive-size GOP is used to execute MCTF. The experimental results indicate that the proposed scheme can improve coding efficiency greatly comparing to the fixed-size GOP.