论文部分内容阅读
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。
The intra-prediction mode selection problem is a bottleneck problem in H.264 intra-coding. The reference algorithm traverses the 9 prediction modes of each 4 × 4 sub-block to find the best mode, which consumes a great amount of computation and is not conducive to real-time coding. This paper presents a fast intra prediction 4 × 4 sub-block mode selection algorithm. The algorithm uses sub-sampling model and boundary texture information extraction technology, according to the principle of “good enough to stop”, stop the algorithm ahead of schedule, reduce the number of candidate modes, simplify the mode selection process and reduce the complexity of mode selection. The experimental results show that compared with the pattern traversal algorithm in H.264 reference code, the computational complexity of intra-mode selection is reduced by 56% -78% for all I-frame sequences. Because different sequences have different contents, different QP Parameters, the part of the code running time savings (1% ~ 30%), while ensuring the image quality degradation within acceptable range.