Page 138 - ITU Journal, ICT Discoveries, Volume 3, No. 1, June 2020 Special issue: The future of video and immersive media

ITU Journal: ICT Discoveries, Vol. 3(1), June 2020

9 95
7 90
MOS 5 VMAF 85

3 80
1 75
0 10,000 20,000 30,000 0 5,000 10,000 15,000
Bit rate (kbps) Bit rate (kbps)
Fig. 9 – RD-curves for sequence A2 using
H.265/HEVC as base and anchor codec
9

7
FoodMarket4 MOS 5
42
3
40
PSNR (dB) 38 1 0 5,000 10,000 15,000
Bit rate (kbps)
36
34 Fig. 11 – RD-curves for sequence A4 using
0 5,000 10,000 15,000 H.265/HEVC as base and anchor codec
Bit rate (kbps) Results show that LCEVC provides a significant
improvement over the respective anchor codecs, in
100 terms of both objective and subjective metrics.
90 In addition to the previous experiment, the video
VMAF 80 data set A has been encoded using the reference

implementations of LCEVC, H.264/AVC and
70 H.265/HEVC at higher bit rates. Namely, the
0 5,000 10,000 15,000 following four QPs have been used to encode the
Bit rate (kbps) anchors: 27, 32, 37 and 42. The resulting BD-rates
are reported in Table 3. By way of example, Fig. 12
10 highlights the RD-curves for objective metrics of the
8 sequence ‘ParkRunning3’ over the extended bit rate
MOS 6 range for H.264/AVC and LCEVC over H.264/AVC.
When comparing Table 3 with Table 2 and Fig. 12
4
with Fig. 5, it can be seen that the performance is
2 consistent also across an extended bit rate range.
0 5,000 10,000 15,000
Bit rate (kbps)
Fig. 10 – RD-curves for sequence A3 using Table 3 – Coding performance comparison of LCEVC
H.265/HEVC as base and anchor codec (LTM 4.0) over AVC/HEVC anchor (JM/HM) –
extended bit rate range
Video data set PSNR VMAF
Fortnite (Part 1) (base & anchor codec)
39
Avg. A (H.264/AVC) −31.97% −42.80%
PSNR (dB) 37 Avg. A (H.265/HEVC) −5.45% −23.55%

35
33
0 5,000 10,000 15,000
Bit rate (kbps)

116 © International Telecommunication Union, 2020