基于混合特征的H.264/AVC快速帧内预测模式决策

H.264/AVC中采用率失真优化(RDO)代价函数对多种帧内预测模式进行全搜索决策,为了克服这种决策方法运算复杂度较大的弊端,提出了一种基于混合特征的快速帧内预测模式决策算法。该方法针对最耗时的4×4亮度块模式决策环节,强调了DC模式的重要性,并将其与混合代价函数特征相结合形成混合决策信息,以此对候选模式进行预先筛选,从而加快了决策速度。实验结果表明,该方法可以使图像序列的I帧编码速度提高近30%,且与原全搜索算法相比,其峰值信噪比(PSNR)的减少和比特率的增加均小到可以忽略不计。

Mixed features-based fast intra-prediction mode decision for H. 264/AVC

Due to the use of Rate Distortion Optimization(RDO) for the decision of multiple intra-prediction modes, a full search method used in reference software of H.264/AVC comes at a cost in greatly increased complexity at the encoder. In order to improve the time efficiency of the encoder, a mixed features-based on fast intra-prediction mode decision method aiming at 44 luminance blocks was proposed. DC mode was emphasized in this method, and it was combined with the mixed features of cost functions to form mixed decision information. Advance modes were filtered by virtue of these mixed features, and the decision approach was speeded up. The experiment results show that the new scheme is able to achieve a reduction of up to 30% encoding time for I frames of image sequences, with negligible Peak Signal-Noise Ratio (PSNR) loss and bit rate increase compared with the original H.264/AVC reference software.