MSU HEVC, H.264, VP9 and AV1 Codec Comparison 2018

[IgorC]

http://www.compression.ru/video/codec_comparison/hevc_2018/

It's not finished yet but there are some partial results
They have included VMAF metrics this time.


There is also comparison from last year (2017)
http://www.compression.ru/video/codec_comparison/hevc_2017/

VP9 is in the middle between x265 Veryslow and Placebo
AV1 is significantly better than both but unusable in real cases (extremely slow)

(page 15)
http://www.compression.ru/video/codec_comparison/hevc_2017/MSU_HEVC_comparison_2017_P5_HQ_encoders.pdf

[iwod]

I think the 2018 is much more meaningful then the 2017 one, VMAF is much more useful then PSNR.

I hope Netflix will continue to improve VMAF model.

[IgorC]

I think the 2018 is much more meaningful then the 2017 one, VMAF is much more useful then PSNR.

Well, they used SSIM in 2017 which correlates considerably better than PSNR.
VMAF is great though it has some flaws as well.

[hajj_3]

I think the 2018 is much more meaningful then the 2017 one, VMAF is much more useful then PSNR.

I hope Netflix will continue to improve VMAF model.

That and the fact that AV1 will be in the comparison.