一种面向新型视频编码器的快速算法

为降低多功能视频编码标准(VVC)编码的复杂度,提出一种面向VVC的帧内快速编码算法.首先,根据视频内容的时空域相关性,使用反向传播(BP)神经网络对CU的划分深度进行预测;然后,使用统计概率对CU的划分模式进行选择;最后,编码时跳过不必要的划分模式以节省编码时间.实验结果表明,与原始编码器相比,该算法平均可节省59.82%的编码时间,且在同等编码质量情况下比特率的平均增加值(BDBR)仅为2.05%.

A fast coding algorithm for new video encoder

Versatile video coding (VVC) standard improves the coding efficiency while increasing the coding computational complexity. According to the content characteristics, the video image will be divided into different sizes of blocks (coding unit (CU)) for better compression efficiency in VVC. The process of determining the best video image partition mode takes a lot of the encoding time. In order to reduce the coding computational complexity of VVC, a fast intra coding algorithm is proposed in this paper. First, based on the spatial-temporal coherence of the video content, a back propagation (BP) neural network is built to make CU depth prediction. Then, we select the partition modes of CU based on statistical probability. Finally, the unnecessary partition modes are skipped to reduce the encoding time of the encoder. Experimental results show that the proposed algorithm can save 59.82% encoding time with 2.05% Bj?ntegaard delta bit rate (BDBR) on average compared to the original encoder.