JVET-Y0047 AHG4: First report of spatial scalability verification tests [P. de Lagrange, F. Urban, E. François (InterDigital), W. Hamidouche (INSA)]

This contribution reports the results a formal visual quality assessment of VVC in spatial scalability mode, following preparation work described in JVET-X0202. With equal bit rate for base HD and enhancement UHD layer, PSNR-Y based performance is ‑24% vs simulcast, while MOS-based performance is around ‑15%; other layer bit rate allocations are tested in JVET-Y0048. Part of those results were used as a subjective evaluation report of VVC scalable profile in response to the TV 3.0 project Call for Proposals from the Brazilian Digital Terrestrial TV Forum (see JVET-U0128, JVET-V0167 and JVET-X0202). Beyond that, these tests can be used as a basis for JVET verification tests of VVC in scalable mode.

2x spatial scalability is investigated.

Performance comparison vs. simulcast and single-layer

Unlike SHVC verification test, which used 0.5 s IRAP period, 1 s was used

MCTF preprocessing was not found effective in scalable coding

Base layer and low res in simulcast almost identical

Quality/rate of base vs. enhancement layer was equivalent to SHVC verification test.

From sequences that were previously used in SHVC verification tests, it is concluded that gain against simulcast are similar (with PSNR based criteria).

Base layer quality might not be chosen in an optimum way over the different rate points. needs to be improved. It is also suggested that one possibility would be including an upsampled base layer to determine appropriate settings. Cases should be investigated where scalability has demonstratable benefit. Also, it should be avoided that quality saturates at the high rate end.

The base layer should be viewable.

At the highest rate points, saturation occurs.

Should a verification test include SHVC, or would it be sufficient to demonstrate the benefit vs. simulcast?

It was suggested to further investigate the definition of a scalable VVC verification test, in particular with regard to the optimum rate/quality settings for base and enhancement layers. Put this as mandate of AHG4.

JVET-Y0048 is also related to this.

Test material (3)

Contributions in this area were discussed in session 13 at 1445–1520 UTC on Monday 17 Jan. 2022 (chaired by JRO) and in session 24 at 0010–0030 UTC on Friday 21 Jan. 2022 (chaired by JRO).

JVET-Y0042 (section 4.10) is also related.