JVET-S0246 Results of dry run subjective assessment of SDR UHD verification test [V. Baroncini, M. Wien]

The draft plan for VVC verification testing (Draft 2) [JVET-R2009] included a set of UHD SDR test sequences under consideration for use in the formal verification tests of VVC. According to the timeline of this plan, a preselection and preparation of test sequences and corresponding rate points for viewing has been carried out as described in document JVET-S0041. Out of the 13 test sequences listed in JVET-R2009, 8 test sequences were selected. The selection process and the experts viewing session subsequently conducted are described in this document. Starting from the selection described in JVET-S0041, a set of 4 test sequences were selected to run a Dry Run formal subjective assessment. A set of 5 rate points was selected for the assessment of each test sequence. The results indicate that the selected rate points cover the intended MOS range for most of the test sequences. Modifications and improvements to the set of test sequences and to the VTM and HM QP selection are suggested.

This was presented Tue 06-23 1455 UTC (chaired by JRO).

Some alignment of rate points for Neptune Fountain, where VTM saturates at highest rate. More intermediate points, and/or slightly lower lowest rate point.

It is also observed that in many cases where VTM is slightly worse in PSNR, it is slightly better in MOS, which seems to indicate that subjective benefit over HEVC is larger than PSNR benefit.

It is also suggested that a fifth sequence (Tall Buildings) should be used in the final verification test.