基于自适应码率分配的压缩传感深度视频编码方法  

Coding Scheme for Compressive Sensing Depth Video Based on Adaptive Bits Allocation

作　　者：王康 兰旭光[1] 李翔伟 WANG Kang1 , LAN Xuguang1 , LI Xiangwei2(1. Institute of Artificial Intelligence and Robotics, Xi＇an Jiaotong University, Xi＇an 710049; 2. Xi＇an Institute of Optics and Precision Mechanics, Chinese Academy of Sciences, Xi＇an 71004)

机构地区：[1]西安交通大学人工智能与机器人研究所,西安710049 [2]中国科学院西安光学精密机械研究所,西安710049

出　　处：《模式识别与人工智能》2018年第4期293-299,共7页Pattern Recognition and Artificial Intelligence

基　　金：国家自然科学基金委员会共融机器人计划重点项目(No.91748208);国家自然科学基金项目(No.61573268);国家重点研发计划(No.2016YFB1000903)资助~~

摘　　要：压缩传感深度视频(CSDV)是由深度视频经过压缩得到,它的冗余信息仍然巨大,由此,文中提出基于高斯混合模型和边缘码率分配的深度视频编码方法.在时域方向上,使用压缩传感,压缩八帧深度视频,得到一帧CSDV图像.为了减小量化的计算复杂度,将一帧CSDV图像分割成一系列大小相同且互不重合的视频块,使用Canny算子作为边界提取工具提取视频块的边界.根据每个视频块中非零像素所占的百分比,给不同的视频块分配不同的比特数.在模型中,使用高斯混合模型建模这些视频块,用于设计乘积矢量量化器,再使用乘积矢量量化器量化这些视频块.By utilizing the compressive sensing in the depth video,the compressive sensing depth video（ CSDV） is obtained. However,the redundancy of CSDV is still huge. A coding scheme for compressive sensing depth video（ CSDV） based on Gaussian mixture models（ GMM） and object edges is proposed. Firstly,the compressive sensing（ CS） is utilized to compress 8 depth frames to acquire a CSDV frame in the temporal direction. A whole CSDV frame is divided into a set of non-overlap patches,and object edges in the patches are detected by Canny operator to reduce the computational complexity of quantization. Then,variable bits for different patches are allocated based on the percentages of non-zero pixels in every patch. The GMM is employed to model the CSDV frame patches and design product vector quantizers to quantize CSDV frames.

关 键 词：高斯混合模型 深度视频编码 压缩传感 视点合成 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]