摘要
在利用宏块分割进行边界方向检测判断图像纹理方向的帧内预测算法的基础上提出了一种改进算法。首先根据宏块平坦程度对宏块预测方式预判——4×4块或16×16块,然后在4×4块预测方式中结合相邻方向模式的相关性进一步减少预测模式数,在16×16块预测方式中直接采用边界方向检测的方法选取预测模式,最后对8×8块运用类似16×16块的方法进行预测。实验结果表明,该算法与原文算法相比,编码时间减少了18%左右,峰值信噪比(PSNR)和输出码率基本保持不变。
The improved algorithm of intra prediction for H.264/AVC based on the direction of image texture was presented by testing the direction of the boundary in the divided macro block. Firstly, the intra prediction type was decided by the smoothness which a macro block has. Secondly, the prediction combined by the correlation of neighbor direction prediction mode in the intra 4×4 prediction mode to reduce the number of prediction mode was made, or used the intra 16×16 directly to algorithm. The method of testing the direction of the boundary was used to choose the mode. At last, the intra 8×8 was predicted just like that in the intra 16×16. Experimental results show that, this algorithm compared with the original algorithm, can decrease the coding time by 18%, while the loss in Peak Signal-to-Noise Ratio is negligible and the out video bit rate increases slightly.
出处
《辽宁工业大学学报（自然科学版）》
2013年第4期235-237,241,共4页
Journal of Liaoning University of Technology(Natural Science Edition)
基金
辽宁省科技厅攻关项目(2008217005)
辽宁省教育厅项目(2009T049)
关键词
H
264
AVC预测模式相关性
图像纹理方向
帧内预测
H.264/AVC
the correlation of prediction mode
texture direction of image
intra prediction
