I have a 720p, 30 fps, 8-bit source video. I'm downscaling it to 480p, and I want to measure quality using PSNR/SSIM under two scenarios:
Keeping crf/speed settings same, I want to encode in 8-bit(yuv420p) and 10-bit(yuv420p10le). I want to measure SSIM/PSNR using ffmpeg's filter. Now normally I'd do something like
ffmpeg -i reference.mkv -i distorted.8bit.mkv/distorted.10bit.mkv -lavfi "ssim" -f null -
where
reference.mkv
is 480p lossless ffv1 from the same source. Now the problem here is, thatreference.mkv
is 8-bit but one of the two distorted videos is 10-bit. Can we compare a 10-bit distorted video with an 8-bit reference?I create two videos, with the same crf/speed/bit-depth settings. The first has the same fps as the source (30 fps). The second is downsampled to 18 fps. I want to measure the deterioration in quality (if any) caused purely by downsampling the fps.
Again, the problem here will be that the reference lossless ffv1 is 30 fps, but one of the distorted videos is 18 fps. How do we compare objective metrics for two videos that differ in frame-rates?