360° video is a relatively new video format. The whole scene around a single point is recorded. During playback the user can choose and change his viewing direction. Interest in 360° video and cameras started growing in mid 2015 and it was of particular interest in the development of the standard Versatile Video Coding (VVC), which was approved in late August 2020. This work was performed in parallel to the development of VVC and focused on the development of coding tools specific to 360° video. The first developed tool allows the introduction of uncoded areas into the pictures of a coded video. In this work it is shown how uncoded areas alone can be applied in order too efficiently encode 360° video represented by cube faces in a non-compact packing, outperforming coding in the compact packing. Further, the uncoded areas are used as a building block of the other tools developed in this thesis. The main contribution of this work is a method of improved motion compensation for 360° video in cubemap representation, called geometry padding. It works by correcting geometrical distortions which occur at cube face boundaries. Geometry padding improves the prediction signal used by motion compensation and thus increases the coding performance. It also mitigates otherwise clearly visible artifacts. 360° video also poses problems for the conventional loop filters used in video coding. In particular, application of the traditional deblocking filter (DBF) at face boundaries in 360° video in cubemap representation can lead to the creation of additional artifacts. This issue was addressed with the development of a modified DBF.

Hersteller:
Shaker Verlag GmbH
info@shaker.de
Am Langen Graben 15a
DE 52353 Düren