一种面向H.264/AVC的快速帧内预测选择算法

为提高I帧的编码效率,最新的视频压缩编码标准H.264/AVC,在空间域上 从多个方向进行多种模式的帧内预测,并使用率失真优化算法选择最佳模式,算法复杂度很 高.为此,提出了一种快速的帧内预测算法.针对亮度块的模式选择过程,提出了基于图像 熵值和滑动窗口机制自适应地调整熵值阈值的方法对宏块模式进行预判;针对亮度模式中多 个预测方向的选择,充分利用中间结果,预先排除一些可能性小的预测方向;在亮度块和色 度块相结合的率失真统计中,改变算法结构,预先计算色度块最佳模式.实验结果表明,算 法在不降低图像质量的基础上,编码速度平均提高76%,码率平均增加2.6%.

Fast selective algorithm of Intra prediction for H.264/AVC

The newest video coding standard H.264/AVC provides several predicting modes and directions for Intra prediction in the spatial region,and the optimal mode was selected by rate-distortion optimization(RDO) with high coding complexity.In order to solve this problem,the fast selection algorithm was proposed from different aspects: Firstly,for luminance mode selection,the method was proposed that utilizes the image entropy value to predict the mode of the macroblock,and updates the threshold adaptively by a sliding window.Secondly,intermediate computing results were reused to reduce the predicting directions.Finally,in the RDO performance,the arithmetic structure of Intra prediction was changed with calculating the best chromance mode separately.Experimental results show that with similar distortion performance,the algorithm that adopt the above three strategy can efficiently reduce the computational complexity up to 76% with 2.6% increase of bits rate averagely.