D.2 Rendered viewports

26.1183GPPRelease 17TSVirtual Reality (VR) profiles for streaming applications

Error! Reference source not found. illustrates an example of clustering and the associated viewports. The first three evaluated viewports are all with the distance D (indicated by the blue circle), and are thus assigned to the same cluster. Note that the cluster center moves a bit for each new viewport which is added to the cluster.

Viewport #4 is too far away from the center of cluster #1, and thus starts a new cluster, which eventually gathers three viewport members. Then viewport #7 is too far away from the center of cluster #2, and again starts a new cluster.

Figure D.2-1: Clustering example

For each cluster j, the final averaged viewport parameters can be derived as follows, assuming there are N viewports in the j:th cluster. Note that the center azimuth and tilt averaging also needs to handle the special case around -180/180 degrees, as some values might be positive (e.g. 176 degrees), while others might be negative (e.g. -178 degrees). This special case is not shown in the equations below.

Note also that the azimuth and elevation range (i.e. the visible coverage of the viewport) might often be the same for every viewport, unless the user explicitly changes the field-of-view for the device. For consistency, and to catch any during-session field-of-view changes, these two parameters should still be averaged.

Figure D.2-2 below illustrates an example of the duration filtering. The user starts by looking at the upper left part of the media (viewports #1 to #3), then make a very brief glance to the right (viewport #4), and then moves back to the upper-left again (viewports #5 and #6). Then the user moves his gaze to the lower-right part (viewports #7 to #10).

Assume here that the duration T is set to 4 times the value of the viewport sample rate X, i.e. a cluster needs to have a duration corresponding to at least four viewports to be reported. Here four clusters are formed, but before filtering only cluster #4 would be reported. After filtering, clusters #1 and #3 are close enough both in time and distance to add to each other’s aggregated duration, so each of them will be assigned an aggregated duration of 5, and thus be reported. Cluster #2, the quick glance up to the right, has too short duration and will not be reported.

Figure D.2-2 Duration filtering example

Annex E (informative):
Change history

Change history

Date

Meeting

TDoc

CR

Rev

Cat

Subject/Comment

New version

2018-06

SA#80

SP-180270

Presented to TSG SA#80 (for information)

1.0.0

2018-09

SA#81

SP-180647

Presented to TSG SA#81 (for approval)

2.0.0

2018-09

SA#81

Approved at TSG SA#81

15.0.0

2018-12

SA#82

SP-180967

0001

2

F

Corrections to 26.118

15.1.0

2019-09

SA#85

SP-190649

0002

F

Correction of figure references

15.2.0

2020-03

SA#87-e

SP-200038

0003

B

Addition of Feature

16.0.0

2020-03

SA#87-e

Editorial Correction in Change History, Copyright etc

16.0.1

2020-03

Post SA#87-e

Minor Editorial changes

16.0.2

2020-12

SA#90-e

SP-200932

0005

A

Corrections to Video Operation Points

16.1.0

2021-01

Post SA#90-e

Update of History Table

16.1.1

2021-04

SA#91-e

SP-210038

0006

4

B

Operation Points for 8K VR 360 Video

16.2.0

2021-04

SA#91-e

This version is identical to 16.1.1 and to be used as the latest Rel 16 version.CR0006r4 had to be implemented on 16.1.1 to build Rel 17 version (instead,16.2.0 has been inadvertently updated with CR0006r4 and hence this rollback was necessary).

16.2.1

2021-06

SA#92-e

SP-210532

0009

F

Spatial positioning of the chroma samples for VR Video Profiles

16.3.0

2021-06

SA#92-e

SP-210534

0007

2

B

Operation Points for 8K VR 360 Video

17.0.0

2021-06

SA#92-e

SP-210534

0008

C

Addition of HLG transfer characteristics

17.0.0

2021-06

SA#92-e

SP-210533

0010

1

F

Spatial positioning of the chroma samples for VR Video Profiles

17.0.0