I noticed that a common metrics for image quality assessment (Monte Carlo path tracing) is MSE between the generated image and a reference one. My question is: why isn't the standard deviation ever mentioned? Isn't it necessary? Shouldn't we generate n number of samples, compare them all with the reference, and then report the average and the standard deviation? This is how I have always seen results in technical papers, except for papers in Graphics. I am wondering if the fact that path tracing is unbiased plays a role in the missing std dev. Thank you.

  • 1
    $\begingroup$ It's very hard to get unbiased estimation of the standard deviation, not so for variance, where Bessel's correction provides an unbiased estimate. $\endgroup$
    – lightxbulb
    Sep 22 '19 at 17:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.