Some background of my scenario: I have two images of indirect radiance, both are from an animation. I will reproject an pixel from the previous frame to current frame, and I want to do comparisons with them. Both have been path traced with quite high sample count, but some noise still remains.

I want to focus on reprojecting, so two questions:

  1. Does it make sense applying a denoising filter to the images prior to comparing them, to lower the variance? Frankly, I don't see a reason why not, but not completely sure.
  2. What filter would give the highest quality (PSNR/SSIM against higher spp count) results? I have stumbled upon gaussian, blackman-harris and bilateral filters, but they have been used on final (indirect + direct) colors.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.