You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In quat/visual/fullref.py the PSNR and SSIM calculation bitdepth argument is fixed to 10. I think this results in invalid psnr output for 8 bit videos because I saw PSNR values of around 100 when I think they should be around 30 to 50 righ?
Edit: I changed the bitdepth of psnr calculation to 8 and still the values are around 100, why is that?
The text was updated successfully, but these errors were encountered:
well, the reason why it is fixed to 10 bit is that the maximum bitdepth for the videos which have been used during development and evaluation was 10 bit, therefore this was to be assumed to be the maximum,
however, the iterate by frame methods use from skimage import img_as_uint as an aggregation and unification step, thus each frame will be automatically a 16 bit uint frame. (the skvideo.measure.psnr method seems to convert each frame anyhow later to a float32 for the calculation of psnr), this seems to be a bit mystical.
In
quat/visual/fullref.py
the PSNR and SSIM calculationbitdepth
argument is fixed to 10. I think this results in invalid psnr output for 8 bit videos because I saw PSNR values of around 100 when I think they should be around 30 to 50 righ?Edit: I changed the bitdepth of psnr calculation to 8 and still the values are around 100, why is that?
The text was updated successfully, but these errors were encountered: