基于纹理特征的H.264/AVC帧内预测模式快速选择算法

H.264/AVC视频标准中的帧内编码通常采用率失真优化技术,通过遍历所有可能的编码模式,选择其中一个具有最小率失真代价的模式作为最佳帧内预测编码。率失真优化技术在提高编码可靠性、准确性、抗误码率的同时,增加了编码的复杂度和搜索时间。在分析标准帧内预测的情况下,计算当前4×4块和重建4×4块的残差值,若小于某一阈值可提前结束宏块编码,反之则利用图像块的纹理方向与预测模式选择之间的相关性,提出一种快速的帧内4×4块预测模式选择算法。该算法在结合改进的Canny算子和平方梯度法的基础上,提取出当前块的纹理方向,并详细分析了纹理方向与预测模式之间的对应关系。实验仿真在H.264/AVC的参考模型JM13.2平台上完成。结果表明,与标准算法相比,该算法在编码码率持平和峰值信噪比保持不变的前提下,编码时间有较大幅度的下降。

Fast intra-prediction mode selection algorithm for H.264/AVC based on texture

Intra-frame encoding in H.264/AVC video standard usually adopts Rate Distortion Optimization (RDO) technique. By computing all possible coding modes, the one which has the lowest cost of rate distortion is selected as the best mode. The RDO technique improves the reliability, accuracy and error resilience of encoding while increasing the complexity and search time of the encoding. The residual value between the current and reconstructed 4×4 block is calculated in the circumstances of analyzing the standard intra-prediction. If the value is smaller than a certain threshold the macro block coding can be ended in advance. Conversely, a fast 4×4 intra-prediction mode selection algorithm was proposed by utilizing the correlation between the texture direction of block and the choice of predict modes. Based on the improved Canny operator and square gradient method, this algorithm extracts the texture of the current block and analyzes the corresponding relations between texture direction and the predict modes in detail. Experiments were carried out on the reference model JM13.2 of H.264/AVC. The results show that, compared with the standard algorithm, the encoding time is substantially declined in the case of keeping the bit rate and the PSNR are nearly unchanged.