JVET-AN0189 AHG17/AHG10: On perceptual coding for next-generation video coding standard [Y. Zhao, A. Karabutov, T. Guo, P. Jia, E. Alshina (Huawei)]

See section 4.16.1.

JVET-AN0189 AHG17/AHG10: On perceptual coding for next-generation video coding standard [Y. Zhao, A. Karabutov, T. Guo, P. Jia, E. Alshina (Huawei)]

NNVC and ECM have been developed with MSE-oriented optimization and evaluated with PSNR-BD-rate. It is suggested to consider perceptually oriented coding tools in the development of next-generation video coding standard.

In a future standards development, usage of more perceptual metrics than MS-SSIM is probably welcome in particular for evaluation. It might be dangerous to use for optimization. Ways must also be found to agree on a solution where results are persistent.

Also contributions on tools with perceptual benefit and perceptual encoder optimization would be welcome.

It was commented that the visual quality also highly depends on target displays.

It was commented that for generative approaches, also metrics like MSSIM and VMAF might not work.

Often (see example HDR) metrics contradict each other.

In the context of CfE responses, the correlation between visual and results and MS-SSIM/PSNR should be investigated.

Further study in AHG4 and AHG17.