Regional segmentation based Low Complexity Illumination Compensation Method in Stereo Vision
Illumination Compensation Histogram Matching Low-complexity 3DTV Stereo Vision;
- 원문 URL
This study proposes a low-complexity Illumination Compensation (IC) scheme for the Stereo Vision. In 3D Video coding, there are some mismatches such as Illumination mismatch, focus mismatch, etc. The illumination mismatches decrease PSNR and bring about optical fatigue when people watch video. To solve this problem, it has to compensate illumination in preprocessing step which is before the encoding step. There are some IC algorithm For example, Histogram mapping(HM) method, Time-constant HM method, Macroblock(MB) based affine model algorithm and feature based color correction method. However, each algorithm has some problems. The algorithm, that uses macro-block, can cause serious blocking effect and weak at occlusion area. The HM method and Time-Constant HM method provide low-complexity and less operations. On the other hand, HM method cannot consider occlusion area and Time-Constant HM method can cause mismatch between views. The feature-based color correction method is appropriate for performance like PSNR and bitrate, but it has complex operations. In the proposed IC scheme, it selects adaptively two IC schemes through using the movement between frames and compensates each separated area. On the same frame, if there is area where the movement appears, Time-Constant HM method is used at only this area since it has high possibility of occlusion area. At the same time, if there is the area where movement is stable, HM method is used at the stable area, since it is more robust. It uses combined proposed IC scheme to increase 0.08dB-1.4dB PSNR performance compared to HM method and decrease 95% operation complexity compared with Feature-based color correction algorithm.