一种深度图帧内预测模式决策改进算法

为减小3D-高效视频编码(three dimensional-high efficiency video coding,3D-HEVC)的编码复杂度,提出一种深度图帧内预测模式决策改进算法.一方面,采用拉普拉斯边缘检测法对是否遍历深度模型模式(depth modeling modes,DMMs)进行快速判决;另一方面,利用预测单元(prediction unit,PU)的楔形分割线与其纹理特征的相关性,只对候选预测模式中的帧内角度模式相关的楔形分割进行搜索,减少楔形分割模式遍历的数量,实现DMM1模式的快速决策.经测试,算法在平均编码比特率增加很少的情况下,深度模型模式的平均编码时间降低了53.65％,而深度图的合成质量基本不变.

Improved depth intra mode selection algorithm

In order to reduce the 3D HEVC coding complexity, there is an improved intra prediction mode decision algorithm for depth maps. On the one hand, the proposed algorithm uses Laplace operator edge detection, and determines whether it traverses depth modeling modes or not. On the other hand, a fast DMM1 decision can be realized efficiently by using the correlation between the line of wedgelet pattern and its textural features in prediction unit. Then, it reduces the number of wedgelet pattern modes through wedgelet pattern related to angle modes in candidate list. After tests, the experiments show that the DMMs algorithm is able to reduce the average time of the encoder by 53.65% while minimally reduced bitrate of coding. Besides, the quality of encoded video almost remains unchanged.