基于能量谱纹理分析的帧内预测算

A Intra Prediction Algorithm Based On Texture Analysis of Power Spectrum

H.264标准采用多种帧内编码模式,编码效率高但同时运算复杂度也高,为降低帧内预测计算复杂度,提出一种新的快速帧内预测算法。先对宏块进行傅里叶变换,将其从空域转换到频域,求出能量谱并提取它的两种度量:环特征和楔特征,这两种特征能描述纹理相关信息,再利用相邻块的空间相关性来减少帧内预测模式数量。实验表明,在保持图像质量和码率大致不变的基础上,极大地降低了计算复杂度,提高了编码速度。

H,264 standard employed several coding modes to obtain a high compression efficiency with lots of computations .To re duce the complexity of intra prediction algorithm,a new fast intra prediction algorithm was proposed.First, the macroblock would trans late from space domain to frequency domain through Fourier transform, then the power spectrum was obtaind and two measures：ring feature and wedge feature were fund.This two features can show the textural features,then taking use of the spatial relevance among the neighbouring blocks to reduce the choices of mode.Experimental results show that the proposed algorithm reduce the computation al complexity of intra prediction coding dramatically and enhance the rate of coding with little loss of PSNR and increment of bit rate.