一种高效的三维视频深度图帧内编码算法

Efficient intra coding algorithm for depth map in 3Dvideo

针对三维视频深度图的压缩问题,提出一种高效的帧内编码算法.该算法根据深度图的典型特征,通过二级变换有效减少了原始深度图的数据量.首先利用深度图由大片均质区域构成以及深度图边缘与对应视频图像边缘呈高相关性的特性,采用超像素分割和赋值策略对原始深度图进行尺寸变换;然后利用深度图灰度等级偏少的特性,采用直方图统计策略对深度图进行灰度变换;最后采用H.264无损帧内编码算法对二级变换后的深度图进行编码以实现对深度图边缘的有效保护.实验结果表明:与现有算法相比,本文算法大幅提高了深度图编码效率和虚拟视点合成质量.

Towards the depth map compression issue in 3Dvideo coding,an efficient depth map intra coding algorithm was proposed.By taking full advantage of the typical feature of the depth maps,the proposed method employed a two-level transform to reduce the data volume of the original depth maps.First,based on the fact that the depth map contained large homogeneous areas and the occurrence of edges in depth map was highly correlated to the corresponding texture image,the original depth map was resized by superpixel segmentation and pixel assignment strategy.Then,based on the feature that gray level of the depth map was far less than 256,the gray level transform was conducted on the resized depth map through histogram method.Finally,in order to protect the edges of the depth map,the two-level transformed depth map was compressed by H.264 lossless intra coding.The experimental results demonstrate that the proposed method can significantly improve the rate-distortion performance of the depth map coding and the subjective quality of the virtual views.