Research On Video Coding-Based Compression Of Point Cloud Sequence

With the increasing capability of scanning devices,the acquisition of highly detailed photo-realistic point clouds becomes available.Thus,dynamic 3D scenes such as persons or objects in motion,can be represented by dynamic point cloud sequences.Unlike 2D images,point clouds provide complete information of the scene(or object)and can be rendered from any viewpoint.Therefore,these sequences are widely used in many emerging immersive applications such as virtual reality and augmented reality,3D video conference and 3D free viewpoint sports replay broadcasting.Unfortunately,traditional compression methods for these sequences are suffering from poor temporal and spatial compression performance.Recently,MPEG has proposed a video-coding-based framework named Test Model Category 2(here after called TMC2)as the reference software to standardize the compression of point cloud sequences.In the software TMC2,each input frame is segmented into patches and projected on 2D images.And then,these 2D images are compressed utilizing standard2 D video coding technologies.Although it achieves significant improvement in coding efficiency compared with traditional compression methods,it is still imperfect and can be improved in specific aspects.In this paper,we propose a temporal and spatial consistent patch packing and coding method based on TMC2.The proposed method divides the input sequences into some groups adaptively,in which patches in all frames are packed in a consistent way.Thus,the generated occupancy maps as well as the corresponding depth and texture images,are highly consistent both in temporal and spatial,which benefits the subsequent 2D video coding greatly.Additionally,redundancies between patch metadata information are also removed successfully.Therefore,the proposed method can achieve a significant improvement in coding efficiency for point cloud sequence.During the experiment,a total of five test sequences were evaluated at five different bit rate targets.Objective experimental results show that the proposed method achieves significant bit rate savings over the reference method TMC2,with up to 4.9% for geometry and 15.7% for texture in terms of BD-BR.And subjective experimental results show that the reconstructed point cloud generated using the proposed method with a higher visual quality.