基于三维重建技术的多视角视频编码算法

A 3D-reconstruction Technology Based Multi-view Video Coding algorithm

对于多视角视频,尤其是多达几十路视频输入的全景多视角电视这一类应用,必须采取高效的编码方法来对海量的原始数据进行压缩和处理。本文提出了一种基于三维重建技术的全景多视角视频编码算法,三维重建技术的核心思想是基于由多幅二维图像来恢复空间物体三维信息。本算法包括了高效的对应基元匹配和预测编码算法,实现了利用先验空间几何知识,参考同一时刻位于不同通道的两帧来对当前帧做出高效的预测,实验表明,本算法有效的降低了预测编码时的计算复杂度,提高了搜索速度。

When multi-view video, especially panoramic multi-view television application with up to dozens of input video channels is concerned, effective coding methods should be adopted to compress and process the high quantity of raw data. In this paper, we pre- sent a multi-view video coding algorithm based on 3D-reconstruction technology, the essence of which is recovering 3D information of dimensional object from several 2D images. The algorithm includes efficient correspondent points matching and prediction coding algorithms. Given transcendental knowledge of spatial geometrical position, effective prediction on current frame referring two frames from different channels at the same moment is achieved. The experimental results show that the coding scheme effectively reduces the computational complexity and increases the searching speed during prediction.