Discrete Cosine Basis Oriented Homogeneous Motion Discovery for 360-Degree Video Coding

Abstract

Motion modeling plays a central role in video compression. This role is even more critical in 360-degree video sequences given the associated enormous amount of data that need be stored and communicated. While the translational motion model employed by modern video coding standards, such as HEVC, is sufficient in most cases, using higher order models is beneficial; for this reason, the upcoming video coding standard, VVC, employs a 4-parameter affine model. Discrete cosine basis has the ability to efficiently model complex motion fields. In this work, we employ discrete cosine basis to model the underlying motion field in 360-degree video frames. In particular, we discover discrete cosine basis oriented homogeneous motion regions over the current frame. Then employ these estimated motion models, guided by their associated applicability regions, to form a prediction for the target frame. Experimental results show a delta bit rate of \\(5.13\\%\\) can be achieved on 360-degree test sequences, over conventional HEVC, if this predicted frame is used as an additional reference frame.