How to generate "transformed scalar contribution"

Practical and theoretical implementation discussion.
Post Reply
dylan
Posts: 7
Joined: Sun Sep 23, 2012 7:16 pm
Location: Cambridge, UK

How to generate "transformed scalar contribution"

Post by dylan » Mon Sep 24, 2012 5:34 pm

According to Kelemen's Metropolis paper, the input I to the Next pseudocode function (in section 6) is the transformed scalar contribution I*(u). According to earlier in the paper, this is I(S(u))/p_S(u). But what is p_S(u)? Underscore denotes subscript.
Just using I as I* gives an incorrect, apparently edge enhanced image, with bright areas remaining only bright at their edges, and regions around bright areas darkened.
Is there a good page to read that explains this somewhere?

Thanks in advance.

dylan
Posts: 7
Joined: Sun Sep 23, 2012 7:16 pm
Location: Cambridge, UK

Re: How to generate "transformed scalar contribution"

Post by dylan » Thu Oct 04, 2012 1:32 pm

Never mind, I discovered that my use of the weights generated by the algorithm was incorrect. I simply multiplied the weight by the sample and accumulated, dividing by the total number of samples at the end, when in fact you should divide by the total weight. That explains why regions with average weight lower than normal would appear too dark.

Post Reply