Method and apparatus for immersive video formatting
Disclosed herein is an immersive video formatting method and apparatus for supporting motion parallax, The immersive video formatting method includes acquiring a basic video at a basic position, acquiring a multiple view video at at least one position different from the basic position, acquiring at least one residual video plus depth (RVD) video using the basic video and the multiple view video, and generating at least one of a packed video plus depth (PVD) video or predetermined metadata using the acquired basic video and the at least one RVD video.
1. An immersive video formatting method, comprising:
determining whether or not a current view is a basic view;
when the view is not the basic view, determining a reference view of the current view among multiple views;
obtaining a residual view of the current view by removing redundancy, between the reference view and the current view, from the current view;
extracting a rectangular data unit which comprising residual data included in the residual view;
generating an atlas by packing the rectangular data unit; and
generating a metadata for the atlas,
wherein the metadata comprises:
source view information identifying the current view from which the rectangular data unit is derived,
priority information used to determine whether or not the current view is the basic view;
shared view information specifying the reference view of the current view,
packed position information specifying a location of the rectangular data unit in the atlas, and
source position information specifying a location of the rectangular data unit in the current view.
2. The immersive video formatting method according to claim 1 , wherein the current view is an omnidirectional video or a planar video generated by projecting the omnidirectional video on a 2D plane.
3. The immersive video formatting method according to claim 1 , wherein determination of whether the current view is the basic view is based on a size of overlapping regions with other views.
4. The immersive video formatting method according to claim 1 , wherein the method further comprises:
determining a number of reference views,
wherein when the number is determined plural, a plurality of the reference views comprises a basic view and an additional view, the additional view being pruned by the basic view before pruning the current view, and
wherein when the number is determined plural, pruning the current view is repeatedly performed by using each of plurality of reference views.
5. The immersive video formatting method according to claim 1 , wherein the residual data comprises a region, which does not overlap with the reference view, in the current view.
6. The immersive video formatting method according to claim 1 , wherein a pixel which a depth value is less than a threshold value in the rectangular data unit is determined as an invalid pixel.
7. A method of synthesizing immersive video, the method, comprising:
decoding an atlas metadata from a bitstream;
determining whether a current view is a basic view or not;
when the current view is not the basic view, determining a reference view of the current view among multiple views;
extracting, based on the metadata, rectangular data units of the current view and the reference view from the atlas; and
reconstructing the current view based on the extracted rectangular data units,
wherein the metadata comprises:
source view information identifying the current view in which the rectangular data unit is included,
priority information used to determine whether the current view is the basic view or not;
shared view information specifying the reference view of the current view,
packed position information specifying a location of the rectangular data unit in the atlas, and
source position information specifying a location of the rectangular data unit in the current view.
8. The method according to 7 , wherein the method further comprises:
determining a number of reference views,
wherein when the number of reference views is determined plural, a plurality of the reference views comprises a basic view and an additional view, and
wherein the additional view is not the basic view and is pruned by the basic view.
9. The method according to claim 7 , wherein, a pixel which a depth value is less than a threshold value in the rectangular data unit is determined as an invalid pixel.