Welcome To Li Li's Home Page
Home | Research | Publication | Links
1. Video-Based Point Cloud Compression
1) Auxiliary-information-based motion vector prediction
 |
Contributions: 1) General 3D-to-2D motion model using the 3D motion and 3D to 2D correspondence 2) Geometry-based motion vector prediction 3) Auxiliary-information-based motion vector prediction 4) Normaltive and non-normative solutions to utilize the derived motion vector |
2) Occupancy-map-based RDO and partition
 |
Contributions: 1) Occupancy-map-based RDO 2) Occupancy-map-based partition Occupied partition: auxiliary-information-based MVP Unoccupied partition: padding |
 |
Contributions: 1) Unoccupied pixels that are occupied: pad those points using the real points in the original point cloud 2) Unoccupied pixels that are unoccupied: pad the residue of the unoccupied pixels using the average pixel of the occupied pixels |
4) Multiview video coding for plenoptic point cloud compression
 |
Contributions: 1) Provide a multiview-video-based framework utilizing the high efficiency of multiview video coding standard 2) Block-based group padding to unify the unoccupied pixels across the attribute direction 3) Occupancy-map-based rate distortion optimization |
2. Rate Control and Bit Allocation
1) Lambda-Domain Rate Control and Bit Allocation for HEVC
 |
Contributions: 1) Complete lambda-domain R-D analysis framework including both R-lambda and D-lambda models 2) Optimal picture level bit allocation 3) Optimal basic unit level bit allocation |
2) Lambda-Domain Rate Control for HEVC Scalable Extension
 R-lambda R-Q |
Contributions: 1) Initial encoding parameters determination for both HEVC temporal, spatial, and quality scalabilites 2) Optimal picture level bit allocation considering inter-picture and inter-layer dependencies 3) Adaptive model parameter updating scheme |
3) Rate Control for 360-Degree Video Compression
CTU Level Weights |
Contributions: 1) Optimal Coding Tree Unit (CTU) level weight for various projection formats 2) Optimal CTU level bit allocation algorithm based on the CTU level weight 3) Optimal CTU-row level weight specified for the ERP format |
4) Rate Control for Video-Based Point Cloud Compression
 |
Contributions: 1) The first rate control work on video-based point cloud compression 2) Video level bit allocation between geometry and attribute videos 3) Occupancy-map-based CTU level bit allocation 4) CTU level model updating |
2. Advanced Motion Model
1) Four-Parameter Affine Motion Model
 |
Contributions: 1) Four parameter affine motion model 2) Two affine motion prediction: advanced affine motion vector prediction and affine model merge 3) Fast gradient-based affine motion estimation 4) Two practical affine motion compensensation tools: one-step sub-pixel interpolation and adaptive-block-size motion compensation |
2) 360-Degree Specific 3-D Translational Motion Model
 |
Contributions: 1) Unified advanced spherical motion model to handle the geometry distortions 2) Local 3-D padding to handle the face boundary discontinuity 3) All these methods are adapt to various projection formats including ERP, CMP, and OHP |
3. Light Field Image Compression
1) Lenslet-Based Light Field Image Compression
 |
Contributions: 1) 2-D hierarchical coding structure with a limited number of reference frames 2) Distance-based reference frame selection 3) Distance-based motion vector scaling 4) Multi-pass optimal bit allocation |
2) Camera-Array Light Field Image Compression
 |
Contributions: 1) Quadtree-based coding structure: four quadrants, 16 GOPs in total 2) Distance-based reference frame selection and motion vector scaling 3) One-pass optimal bit allocation according to the information of the previous GOP |