Abstract not available in document
JVET-H0006 JVET AHG Report: 360 video conversion software development (AHG6) [Y. He, V. Zakharchenko]
This document summarizes activities on 360-degree video content conversion software development between the 7th and the 8th JVET meetings.
The 360Lib-4.0 software package integrated all adoptions about projection format and metrics calculation:
- Metrics:
- (1) WS-PSNR calculation for ACP (JVET-G0088);
- (2) Add codec S-PSNR-NN
- Projection formats and frame packing:
- (3) Adjusted equal-area projection (AEP) (JVET-G0051);
- (4) Equi-angular cubemap (EAC) projection (JVET-G0056);
- (5) Equatorial cylindrical projection (ECP) with padding (JVET-G0074);
- (6) EAP-based segmented sphere projection (SSP) with padding (JVET-G0097);
- (7) Padded ERP (PERP) projection format (JVET-G0098);
- Software:
- (8) Fix for bug ticket #51, #52, #54
360Lib-4.0 related releases and bug fixes were summarized as follows:
- 360Lib-4.0rc1 with support of HM-16.16 and JEM-7.0rc1 was released on Aug. 4th, 2017;
- 360Lib-4.0 with support of HM-16.16 and JEM-7.0 was released on Aug. 23rd, 2017;
- 360Lib-4.0 based HM-16.16 and JEM-7.0 testing results was released on Aug. 23rd, 2017;
- Ticket #56 was submitted after 360Lib-4.0 release. It only affected codec WS-PSNR calculation for ECP projection format. The fix was provided in the development branch (360Lib-4.1-dev).
The 360Lib software is developed using a Subversion repository located at:
https://jvet.hhi.fraunhofer.de/svn/svn_360Lib/
The released version of 360Lib-4.0 can be found at:
https://jvet.hhi.fraunhofer.de/svn/svn_360Lib/tags/360Lib-4.0/
360Lib-4.0 testing results can be found at:
ftp.ient.rwth-aachen.de/testresults/360Lib-4.0
360Lib bug tracker
https://hevc.hhi.fraunhofer.de/trac/jem/newticket?component=360Lib
360Lib-4.0 results are shown as follows
The table below lists the HM-16.16 based coding performance with different projection formats according to 360° video CTC (JVET-G1030) compared to PERP coding.
HM-16.16-360Lib-4.0 testing (HM PERP coding as anchor)
Projection forma | E2E WS-PSNR for all sequences | E2E WS-PSNR for 8K sequences | E2E WS-PSNR for 6K sequences | E2E WS-PSNR for 4K sequences | ||||||||
Y | U | V | Y | U | V | Y | U | V | Y | U | V | |
CMP | −2.4% | −0.4% | −0.5% | −5.1% | −3.1% | −3.6% | 0.0% | 4.1% | 4.6% | 0.5% | −1.5% | −1.2% |
AEP | −4.3% | −3.0% | −3.1% | −4.7% | −3.5% | −3.3% | −3.7% | −1.5% | −2.0% | −4.3% | −4.8% | −4.9% |
COHP | −1.4% | 2.1% | 1.2% | −5.0% | −0.1% | −1.1% | 3.6% | 6.9% | 5.9% | −0.4% | −1.0% | −1.2% |
CISP | −3.3% | 1.2% | 0.4% | −7.5% | −1.5% | −2.6% | 2.2% | 6.8% | 6.2% | −1.9% | −1.9% | −2.2% |
EAP−SSP | −9.5% | −5.4% | −5.7% | −11.9% | −6.4% | −6.5% | −7.0% | −3.0% | −3.8% | −7.6% | −6.9% | −7.2% |
ACP | −9.9% | −5.3% | −5.5% | −12.4% | −6.3% | −6.4% | −7.6% | −3.3% | −3.9% | −6.7% | −6.3% | −6.3% |
RSP | −8.3% | −4.3% | −4.6% | −11.8% | −6.2% | −6.5% | −4.6% | −1.3% | −1.9% | −5.3% | −4.9% | −4.4% |
ECP | −9.2% | −5.1% | −5.7% | −12.3% | −6.8% | −7.2% | −5.2% | −1.3% | −2.6% | −7.8% | −7.6% | −7.4% |
EAC | −9.9% | −5.3% | −5.6% | −12.5% | −6.4% | −6.5% | −8.0% | −3.4% | −3.9% | −6.1% | −6.2% | −6.1% |
The table below compares the JEM-7.0 PERP coding with HM-16.16 PERP coding. JEM has the higher BD rate saving for 6K sequences (−28.1%) compared to 8K (−20.8%) and 4K (−21.3%) sequences because 6K sequences were captured by moving cameras.
JEM-PERP vs HM-PERP coding (HM PERP coding as anchor)
anchor: HM-PERP | SPSNR-NN (End to End) | WS-PSNR (End to End) | ||||
test: JEM-PERP | Y | U | V | Y | U | V |
Trolley | −15.0% | −30.9% | −32.7% | −14.9% | −30.8% | −32.8% |
GasLamp | −20.3% | −44.6% | −41.5% | −20.2% | −44.6% | −41.5% |
Skateboarding_in_lot | −24.4% | −33.0% | −42.9% | −24.4% | −33.0% | −42.9% |
Chairlift | −29.8% | −48.6% | −45.6% | −29.9% | −48.6% | −45.7% |
KiteFlite | −16.5% | −35.3% | −38.6% | −16.4% | −35.2% | −38.5% |
Harbor | −19.0% | −41.7% | −42.2% | −19.0% | −41.7% | −42.2% |
Balboa | −33.1% | −36.5% | −40.9% | −33.1% | −36.5% | −40.9% |
Broadway | −32.9% | −38.5% | −41.5% | −32.9% | −38.5% | −41.5% |
Landing | −23.0% | −15.5% | −17.9% | −23.0% | −15.5% | −17.9% |
BranCastle | −23.3% | −15.5% | −31.6% | −23.3% | −15.4% | −31.5% |
PoleVault | −17.7% | −19.0% | −20.4% | −17.8% | −19.0% | −20.6% |
AerialCity | −24.8% | −44.7% | −31.0% | −24.9% | −44.8% | −30.9% |
Overall | −23.3% | −33.7% | −35.6% | −23.3% | −33.6% | −35.6% |
8K | −20.8% | −39.0% | −40.6% | −20.8% | −39.0% | −40.6% |
6K | −28.1% | −26.5% | −33.0% | −28.1% | −26.5% | −32.9% |
4K | −21.3% | −31.9% | −25.7% | −21.3% | −31.9% | −25.7% |
The AHG recommended to continue software development of the 360Lib software package.