摘要
帧内预测是H.264编码中用来提升帧内编码压缩效率的编码工具,它通过计算和比较多种帧内预测模式下的宏块率失真代价,并选择最优的帧内预测模式来充分消除图像信号的空间冗余,提升视频信号帧内编码的压缩效率,但多种帧内预测模式下的宏块率失真代价计算使帧内编码运算量成倍增长,阻碍了其实际应用。为此,提出一种帧内预测模式判定快速算法,以提升H.264帧内编码的计算效率。该算法包括3个部分:1)加快4 4帧内预测模式判定的两级预测模式判定快速算法;2)加快宏块帧内预测模式判定的亮度分量帧内预测分区类型提前判定算法;3)融合上述2种算法的宏块帧内预测模式快速判定流程。测试结果显示,与H.264标准参考编解码器JM10.1相比,所提算法编码时间节省52%左右,峰值信噪比仅下降0.03 dB,码率增长约为2.5%。由此证明,所提算法能在保持视频压缩效率基本不变的前提下,有效提升视频编码的运算效率。
Intra prediction is an encoding tool in H.264 encoder to improve compression efficiency of intra-frame coding.Through the calculation and comparison of macro-block-rate distortion costs of multiple intra prediction modes,the optimal infra prediction mode is chosen to thoroughly eliminate spatial redundancy of image signal and improve compression efficiency of the intra encoding of video signals,however,the computation of macro-block rate-distortion cost under multi infra prediction modes makes the macro-block-rate distortion cost increasing several times over,thus the application of infra encoding is impeded.A fast infra prediction mode decision algorithm is proposed to improve the calculation efficiency of infra encoding of H.264 encoder.The proposed algorithm includes three parts:(A) a fast two-stage prediction mode decision algorithm to speed up the decision of 4?4 infra prediction mode;(B) a fast algorithm that speeds up the advanced decision of the type of luminance component infra prediction partition in macro-block infra prediction mode decision;(C) a fast decision process for macro-block infra prediction that merge above-mentioned two algorithms together.Testing results show that about 52% of encoding time can be saved by use of the proposed algorithm than by H.264 standard reference encoder JM10.1,and the decreasing of peak S/N ratio is only 0.03 dB and increasing of bit rate is about 2.5%.Thus,it is proved that the proposed algorithm can effectively improve the operation efficiency of video encoding under the premises that the video compression efficiency is kept basically constant.
出处
《电网技术》
EI
CSCD
北大核心
2012年第7期275-280,共6页
Power System Technology