基于Hough变换的高效视频编码标准帧内预测模式选择快速算法

发布时间：2018-05-30 12:36

本文选题：高效视频编码标准 + 帧内预测　；参考：《计算机应用》2014年12期

【摘要】：针对高效视频编码标准(HEVC)中帧内预测模式选择计算复杂度极高的问题,提出一种基于Hough变换的HEVC帧内预测模式选择快速算法,该算法可以减少对35种预测模式的遍历次数。首先,在对预测模式进行粗选(RMD)过程之前对各尺寸预测单元(PU)进行边缘检测与Hough变换;其次,采用直方图对检测出来的直线段方向角正切值进行统计分析;最后,从35种模式中选择出RMD和率失真优化(RDO)过程中较适用的候选模式集,并在VS 2008运行环境下用C++和Open CV开发应用程序。实验结果表明,所提算法在编码性能稍微有影响的情况下(码率平均升高1.02%,峰值信噪比平均下降0.035 d B),编码时间平均减少了23%。所提算法增强了编码器的实时性,适合高分辨率、大尺寸视频应用。
[Abstract]:Aiming at the high computational complexity of intra prediction mode selection in high efficient video coding standard, a fast HEVC intra prediction mode selection algorithm based on Hough transform is proposed, which can reduce the traversal times of 35 prediction modes. Firstly, edge detection and Hough transform of each dimension prediction unit are carried out before the rough selection of prediction mode. Secondly, the detected tangent value of straight line direction angle is statistically analyzed by histogram. The suitable candidate pattern sets in the process of RMD and rate-distortion optimization are selected from 35 modes, and the application programs are developed with C and Open CV in vs 2008 environment. The experimental results show that the proposed algorithm has a slight effect on the coding performance (the average bit rate is increased by 1.02, the peak SNR is reduced by 0.035 dB on average, and the coding time is reduced by 23 parts on average.) The proposed algorithm enhances the real-time performance of the encoder and is suitable for high resolution and large-scale video applications.
【作者单位】：浙江师范大学数理与信息工程学院;
【基金】：国家自然科学基金资助项目(11026226,40805048) 浙江省自然科学基金资助项目(Y1110510)
【分类号】：TN919.81

【参考文献】