摘要
针对 AVS(Audio Video code Standard)帧内预测中选择最优编码模式流程复杂的问题,提出一种快速帧内预测编码的算法。通过分析 AVS 帧内预测编码的原理及其在参考软件 GDM2.1中的代码实现,推导出判定零系数块的阈值。该快速算法根据零系数块的判定阈值条件判定当前块,对于零系数块,直接编码 cbp 元素,提前中止帧内预测编码,避免了后续的复杂流程;对于非零系数块,采用时域 SAD(Sum of Absolute Difference)与频域 SATD(Sum of Absolute Transformed Difference)相联合的快速选择算法,排除了40%~80%的不可能预测模式,对余下的模式采用率失真代价函数的方法选取最佳预测模式。经实验验证,在 PSNR(Peak Sig-nal to Noise Ratio)降低0.25 db 和码率增加2%之内,AVS 编码时间减少了40%以上,大幅度提高了编码效率。
As to the complexity process of choosing best probable encoding mode in AVS intra prediction,we proposed a fast intra predic-tion encoding algorithm.With the analysis of AVS intra prediction encoding theory and its implementation in reference software GDM2.1,we derived the threshold of zero coefficient block determination.The fast algorithm determines whether or not the current block is a zero coeffi-cient block according to its threshold discriminant condition.For zero coefficient block,it encodes cbp element directly and terminates intra prediction encoding in advance,which avoids the consequent complex process.As for the non-zero coefficient blocks,the fast selection algo-rithm associating the time domain SAD and the frequency domain SATD is used,which excludes 40% to 80% impossible prediction modes, and the rate distortion function is applied to the remained prediction modes to select the best prediction mode.It was verified through experi-ment that when PSNR (peak signal-to-noise ratio)decreased within 0.25db and bits rate increased within 2%,the time of AVS encoding de-creased over 40%,and the encoding efficiency was largely improved as well.
出处
《计算机应用与软件》
CSCD
2015年第6期185-187,共3页
Computer Applications and Software
关键词
AVS
帧内预测
零系数块
SAD
SATD
Audio video code standard (AVS) Intra prediction Zero coefficient block Sum of absolute difference (SAD) Sum ofabsolute transformed difference (SATD)