利用纹理和空间相关性的HEVC帧内预测模式选择

发布时间：2018-06-14 11:23

本文选题：高效视频编码(HEVC) + 视频压缩编码　；参考：《中国图象图形学报》2014年08期

【摘要】：目的提出一种结合图像纹理方向和空间相关性的高效视频编码(HEVC)帧内模式选择快速算法,减少粗模式选择过程的候选模式数目,加快帧内模式选择速度。方法首先,使用Sobel算子获得当前预测单元(PU)的纹理方向,利用此纹理方向选取相应的角度预测模式组成粗模式选择的候选模式列表;另外,利用图像的空间相关性,将相邻预测单元的最优帧内预测模式添加到候选模式列表中。结果本文算法将粗模式选择过程的候选模式数目从35种减少到不超过10种,比特率升高0.41%,峰值信噪比下降0.031 4 dB,编码时间缩短了34.9%。结论实验结果显示,在保持编码质量基本不变的情况下,此算法明显减少了粗模式选择过程的预测模式数量,提高了HEVC的编码效率。
[Abstract]:Aim to reduce the number of candidate modes in the coarse mode selection process and speed up the intra-frame mode selection, a fast intra mode selection algorithm for efficient video coding (HEVC) based on texture direction and spatial correlation is proposed. Methods first of all, we use Sobel operator to obtain the texture direction of the current prediction unit (PUU), and select the corresponding angle prediction pattern to form a list of candidate patterns for rough pattern selection. In addition, the spatial correlation of the image is used. The optimal intra prediction mode of adjacent prediction units is added to the list of candidate modes. Results the algorithm reduces the number of candidate patterns from 35 to less than 10, increases the bit rate by 0.41, decreases the peak signal-to-noise ratio (PSNR) by 0.031 dB, and reduces the coding time by 34.9 dB. Conclusion the experimental results show that the algorithm can reduce the number of prediction patterns in rough pattern selection process and improve the coding efficiency of HEVC.
【作者单位】：合肥工业大学计算机与信息学院;教育部安全关键工业测控技术工程研究中心;合肥工业大学电子科学与应用物理学院;
【基金】：国家自然科学基金项目(61371155) 安徽省科技攻关项目(1301B042023)
【分类号】：TN919.81

【相似文献】