首页 | 官方网站   微博 | 高级检索  
     

简单的4×4块帧内模式选择算法
引用本文:祝敏,陈国斌,郑伟,李东晓.简单的4×4块帧内模式选择算法[J].计算机工程与应用,2006,42(20):44-46.
作者姓名:祝敏  陈国斌  郑伟  李东晓
作者单位:浙江大学信息与电子工程系,杭州,310027
摘    要:帧内预测的模式选择问题是H.264帧内编码中的瓶颈问题。参考算法中对每个4×4子块的9种预测模式进行遍历运算寻找最佳模式,耗费了极大的计算量,不利于实现实时编码。文章提出了一种快速帧内预测4×4子块模式选择算法。该算法采用子抽样模型和边界纹理信息提取技术,按照“足够好就停止”原则,提前中止算法,减小备选模式数量,简化模式选择过程,降低了模式选择的复杂度。实验结果表明:对全I帧序列,与H.264参考代码中模式遍历算法相比较,该算法的帧内模式选择计算复杂度减小56%~78%,由于不同序列内容不同,在不同QP参数下,该部分代码运行时间节约不等(1%~30%),而同时保证图像质量下降在可接受范围之内。

关 键 词:H.264/AVC  4×4模式选择  帧内预测  子抽样  纹理信息
文章编号:1002-8331-(2006)20-0044-03
修稿时间:2006年3月1日

Simple Intra-prediction Mode Selection Algorithm for 4×4 Block in H.264
Zhu Min,Chen Guobin,Zheng Wei,Li Dongxiao.Simple Intra-prediction Mode Selection Algorithm for 4×4 Block in H.264[J].Computer Engineering and Applications,2006,42(20):44-46.
Authors:Zhu Min  Chen Guobin  Zheng Wei  Li Dongxiao
Abstract:Intra prediction mode selection for 4×4 blocks in H.264 is the bottleneck of I-frame encoding.The Full Search(FS) algorithm in H.264 reference software of JVT need to choose one of the 9 prediction modes for each 4×4 block,which results in higher computational load and is not suitable for real-time video communications.In this paper,a fast and simple intra prediction mode selection algorithm for 4×4 blocks is proposed,where an early termination mechanism is further incorporated to avoid unnecessary computations when a good decision can be made in early stage.We first abstract the edge information using texture information of the block,and reduce the number of candidate mode,which simplifies the selection processing and results in lower computational load.A new sub-sampling mode is used in cost function calculation to reduce the number of calculated samples,which leaded to even lower computational load.Simulation results for all-I frame-sequence with different contents show that the proposed algorithm can reduce 56%~78% computational load and 1%~30% time with different QP values in comparison with the Full Search(FS) algorithm in JVT reference software,and remains the image quality acceptable.
Keywords:H  264/AVC
本文献已被 CNKI 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号