Motion Compensated Prediction with Geometry Padding for 360 Video Coding

The paper presents a new geometry padding method for motion compensated prediction in 360 video coding. Unlike the conventional padding method for 2D video coding, which extends samples outside of a picture's boundary by simply copying (repeating) those at the boundaries, the proposed geometry padding method considers the spherical nature of the 360 video and the specific geometry projection format when extending samples outside of a picture's or a face's boundary. The proposed geometry padding method is implemented in the HEVC reference software HM-16.12 for the equirectangular and cubemap projection formats. Simulation results show that, when compared with the conventional 2D padding method in HEVC, geometry padding performs better; the proposed geometry padding gives on average luma (Y) BD-rate reduction of 0.3% for equirectangular projection format coding and 1.0% for cubemap projection format coding all in terms of the spherical PSNR (S-PSNR) metric. The proposed geometry padding is especially effective for high motion sequences, where up to 4.3% BD-rate reduction can be achieved.