【摘要】 随着视频数据的爆炸式增长，视频的存储、传输压力不断加剧。为了进一步提高视频的压缩效率，设计了基于多尺度特征提取的编码预处理算法。该算法可以直接嵌入任何编码器之前，实现即插即用。该算法利用两阶段法，有效衡量了编码器的特点。同时，通过设计多尺度特征提取模块、通道注意力模块，有效提高了视频的压缩效率。实验结果表明，相较于H.266/VVC参考软件VTM10.0，提出的算法以峰值信噪比（PSNR）平均降低0.19 d B为代价，视频压缩效率平均提高19%。更多还原

【Abstract】 With the explosive growth of video data, the pressure on video storage and transmission is increasing. To further improve the compression efficiency of video, this paper designs a coding preprocessing algorithm based on multiscale feature extraction. The algorithm can be directly embedded before any encoder to achieve plug-and-play. The algorithm utilizes a two-stage method to measure the characteristics of encoder effectively. Meanwhile, by designing a multiscale feature extraction module and a channel attention module, the preprocessing model effectively improves the video compression efficiency.更多还原