基于相邻像素的帧内预测算法

Adjacent Pixels-based Intra Prediction Algorithm

为了解决H．264帧内预测中运动物体边缘预测不精确的问题,提出了一种基于块内相邻像素的帧内预测算法．根据块内相邻像素相关性更强的特点,采用层次结构,分阶段预测待编码块内像素．在保证模式方向特性的前提下,充分利用块内已经预测过的像素作为参考,预测其他像素,重新统计最优模式的分布规律,基于统计结果确定该算法的预测模式顺序．实验结果证明,这种基于相邻像素的帧内预测算法可以解决预测不精确的问题,提高帧内编码效率．

In order to solve the problem which pixels located at the boundary of moving objects can not be predicted accurately in H. 264 intra prediction, an adjacent pixels-based intra prediction algorithm is presented in this paper. Because the correlation of pixels in the same block is more closed, some pixels are taken as references to predict other pixels in the same block. Furthermore, new order of prediction modes is deduced according to the novel intra prediction algorithm. The test results indicate that, the adjacent pixels-based intra prediction algorithm can solve the problem, which is mentioned above and improve coding efficiency remarkably.