一种多目分布式视频编码的边信息生成算法

Side information generation for multi-view distributed video code

针对多目视频传感器网络,提出了一种新的空间边信息生成算法,该算法将体现图像空间信息的三焦点张量引入边信息的计算。三焦点张量的匹配点计算采用了改进的基于SIFT(Scale InvariableFeature Transform)特征描述符的匹配算法,在保持对匹配有旋转、尺度缩放、亮度变化不变性的优点同时,大大提高了计算速度。仿真结果表明,无论从率失真表现还是主观质量表现角度进行比较,该算法明显优于常用的边信息生成算法、基于运动的外推法MC-E和基于运动的内插法MC-I。

A novel and efficient side-information frame generator of video sensor network is proposed in this paper, which is based on trifocal tensor. An improved SIFT （scale invariable feature transform） descriptor matching method is used to compute trifocal tensor, it performs much faster with keeping its original advantage： invariable to rotation, multi-scaling and brightness. Simulation results show that the proposed techniques provide potential rate-distortion performance advantages to the MC-E and MC-I method visual performance is also improved.