-
Notifications
You must be signed in to change notification settings - Fork 757
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Wrong PSNR and VMAF values in FFmpeg 5.1.1 #1109
Comments
It is very possible that you have a PTS misalignment, you can try using setpts to align the streams as explained here: https://github.com/Netflix/vmaf/blob/master/resource/doc/ffmpeg.md. To rule out a libvmaf frame mismatch, you can try these sources with the |
|
See #1115 |
We recognized that adding our Wavelet Beam Logo to a video stream didn’t result in a drop of PSNR, when using the PSNR calculation in the FFmpeg VMAF filter. Additionally PSNR is limited at 72dB for 10bpc and to 60 dB for 8bpc (#371 and #1115). This is wrong! "Quantisation noise of the digitisation process" does not matter in the further workflow steps. For distribution encodes we cannot reach this high PSNR values but if you like to use PSNR for other workflow steps, higher PSNR values are important.
Reference video UHD FFv1: black screen with no Wavelet Beam logo:
http://projekte.waveletbeam.com/black.mkv
Video under test: black screen with Wavelet Beam logo:
http://projekte.waveletbeam.com/blackLogo.mkv
This FFMPEG command gives us the correct PSNR value of 28dB:
ffmpeg.exe -i G:\TestFFMPEG5\blackLogo.mkv -i G:\TestFFMPEG5\black.mkv -filter_complex "psnr" -f null -
The VMAF filter with PSNR enabled gives us a PSNR value of 72bB and VMAF of 100:
ffmpeg -i G:\TestFFMPEG5\blackLogo.mkv -i G:\TestFFMPEG5\black.mkv -filter_complex "libvmaf=feature=name=psnr:model=version=vmaf_4k_v0.6.1:log_fmt=xml:log_path="G\:/TestFFMPEG5/black2.xml" " -f null -
We think that 72dB but at the same time 100 VMAF points are wrong results as the difference in the two videos can be easily recognized. Additionally, the PSNR value of 72dB is mathematically wrong.
The text was updated successfully, but these errors were encountered: