arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2205.06754eess.IVcs.CV
收藏

超薄视频编解码器

Slimmable Video Codec

Northwestern Polytechnical University(西北工业大学);Computer Vision Center(计算机视觉中心);Universitat Autonoma de Barcelona(巴塞罗那自治大学);Xidian University(西安电子科技大学);BBC Research & Development(英国广播公司研发部)

Zhaocheng Liu, Luis Herranz, Fei Yang, Saiping Zhang, Shuai Wan, Marta Mrak, Marc Górriz Blanch

摘要:

神经视频压缩已经成为一种结合可训练多层神经网络和机器学习的新范式,实现了有竞争力的率失真(RD)性能,但由于神经架构庞大,内存和计算需求大,仍然不切实际。此外,模型通常针对单个RD权衡进行优化。最近的可精简图像编解码器可以动态地调整其模型容量,以适度地减少存储器和计算需求,而不会损害RD性能。在本文中,我们提出了一个可瘦身的视频编解码器(SlimVC),通过整合一个可瘦身的时间熵模型的可瘦身的自动编码器。尽管一个显着更复杂的架构,我们表明,瘦身仍然是一个强大的机制,以控制速率,内存占用,计算成本和延迟,所有的重要要求,实际的视频压缩。

英文摘要:

Neural video compression has emerged as a novel paradigm combining trainable multilayer neural networks and machine learning, achieving competitive rate-distortion (RD) performances, but still remaining impractical due to heavy neural architectures, with large memory and computational demands. In addition, models are usually optimized for a single RD tradeoff. Recent slimmable image codecs can dynamically adjust their model capacity to gracefully reduce the memory and computation requirements, without harming RD performance. In this paper we propose a slimmable video codec (SlimVC), by integrating a slimmable temporal entropy model in a slimmable autoencoder. Despite a significantly more complex architecture, we show that slimming remains a powerful mechanism to control rate, memory footprint, computational cost and latency, all being important requirements for practical video compression.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑