Inter-picture prediction based on 3D point cloud model

Tong Shao, Dong Liu, Houqiang Li

Abstract: The explosive growth of images in the cloud calls for more efficient image compression that exploits redundancy between images. Traditional solutions to image set compression are sensitive to the diversity in the clustered image sets in the cloud. In this paper, we propose a more efficient inter-picture prediction based on 3D point cloud model. Leveraging the 3D points and camera parameters, we design a patch-wise warping algorithm to generate prediction. Preliminary experimental results show the proposed scheme significantly outperforms the JPEG and HEVC intra as long as high quality reference is provided. The average bitrate reduction is 15.1% in comparison with HEVC intra on the tested 30 images of Notre Dame. Meanwhile, the perceptual quality of the reconstructed images is improved.