learned video compression 论文理解翻译（2）

ML-based video compression.
To our knowledge, the
only pre-existing end-to-end ML-based video compression
approachs are [52, 8, 16]. [52] first encodes key frames, and
proceeds to hierarchically interpolate the frames between
them. [8] designs neural networks for the predictive and
residual coding steps. [16] proposes a variational inference
approach for video compression on 64x64 video samples.
基于机器学习的视频压缩。
据我们所知，现有的基于端到端机器学习的视频压缩方法只有[52,8,16]这3篇文章。文章52，先编码关键帧，然后在关键帧之间分层插如其他帧，文章8为预测编码和残差编码设计了一种神经网络。文章[16] 提出了一种基于64x64视频样本的变分推理视频压缩方法。

Enhancement of traditional coding using ML.
There have been several important contributions demonstrating the effectiveness of replacing or enhancing different components of traditional codecs with counterparts based on neural networks. These include improved motion compensation and interpolation [54, 21, 58, 31], intra-prediction coding [40], post-processing refinement [7, 55, 26, 56, 48, 57,41, 17], and rate control [28].
用机器学习方法提升传统编码方案的性能。
有一些重要的贡献证明了用基于神经网络的对等编码替换或增强传统编码的不同组件的有效性。其中包括改进的运动补偿和插值[54、21、58、31]、帧内预测编码[40]、后处理增强[7、55、26、56、48、57、41、17]和码率控制[28]。