Product quantization of reference frames for fast motion detection at a decoder side
Product quantization of reference frames for fast motion detection at a decoder side
复制标题
DOI：
10.1109/iwait.2018.8369697
复制
发表时间：
2018-05
期刊：
影响因子：
--
通讯作者：
Idomu Fujita;Tomokazu Ishikawa;Yusuke Kameda;I. Matsuda;S. Itoh
中科院分区：
文献类型：
--
作者：
Idomu Fujita;Tomokazu Ishikawa;Yusuke Kameda;I. Matsuda;S. Itoh
最近，探索了基于重建图像之间模板匹配的模板匹配的新运动补偿框架，以进行有效的视频编码。由于模板匹配可以在解码器侧进行，因此不需要传输运动信息。但是，由于解码过程所需的很高的计算复杂性，其应用于实时视频编码实际上受到限制。为了解决此问题，本文提出了使用参考帧的产品量化量量化算法的快速模板。
Recently, a new framework of motion compensated prediction based on template matching between reconstructed images is explored for efficient video coding. Since the template matching can be carried out at a decoder side, it requires no motion information to be transmitted. However, its application to real-time video coding is practically limited due to high computational complexity needed for the decoding process. To cope with this problem, a fast template matching algorithm using product quantization of reference frames is proposed in this paper.