基于对象的水下视频低码率编码算法  

Object based low bit rate underwater video coding algorithm

作　　者：马欢 郁梅[1] MA Huan;YU Mei(Faculty of Information Science and Engineering,Ningbo University,Ningbo Zhejiang 315211,China)

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《激光杂志》2023年第5期84-90,共7页Laser Journal

基　　金：宁波市自然科学基金项目(No.202003N4088)。

摘　　要：由于水下声波信道带宽窄,难以采用高效视频编码(High Efficiency Video Coding,HEVC)实现水下视频低码率传输。提出了一种基于对象的水下视频低码率编码算法。首先对水下对象视频进行时空域下采样以降低其数据量,再采用低延时模式编码少量视频帧作为参考帧。然后,提取水下对象视频非参考帧的特征点,并对特征点和对象掩膜进行编码。在解码端用特征点和掩膜进行对象的粗糙重建,获得对象的初步轮廓和颜色信息。最后,根据粗糙重建对象和参考帧对象的映射关系,采用基于在线学习的方法实现对象的精细重建。实验结果表明,与HEVC相比,所提算法的BDBR-SSIM(Bjontegarrd Delta Bit Rate and Structural Similarity)降低了14.88%。Due to the narrow bandwidth of underwater acoustic channel,it is difficult to transmit underwater video with high efficiency video coding(HEVC)at a very low bit rate.In this paper,an object based low bit rate coding algorithm of underwater video is proposed.Firstly,the underwater video is sampled in time-spatial domain to reduce the amount of data.Then a small number of video frames are encoded by low delay pattern as reference frames;while feature points of underwater video objects in non-reference frames are extracted,and the feature points and masks of the objects are encoded.At the decoding end,the object is roughly reconstructed with the feature points and masks so as to obtain the preliminarily contour and color information of the objects.Finally,according to the mapping relationship between roughly reconstructed objects and objects in reference frames,on-line learning based method is used to realize the fine reconstruction of the objects.Experimental results show that compared with HEVC,the BDBR-SSIM(Bjontegard Delta Bit Rate and Structure Similarity)of the proposed algorithm is reduced by 14.88%.

关 键 词：水下对象视频 低码率编码 特征点 二段式对象重建 声波通信 

分 类 号：TN919[电子电信—通信与信息系统]