The measurement shows the results have higher contrast similarity and structure similarity than those in last blog without ordering step, but lower tone similarity than them.
the weight function
maxin = max(0.5,(Ii/(5*pow(r,1.0f))));
in = min(maxin, 255-Ii);
Tone similarity(not high, better than 'structure-aware ' paper)
Structure similarity(good)
Contrast similarity(good)
Structure similarity(good)
Contrast similarity(good)
I guess there is a bug in the equation ("in" is not used).
ReplyDeleteWhen you normalize, do you usually get a normalization coefficient less than 1? What do you do when the coefficient is higher than 1?
It is always in [0,1]. The normalization is done in two steps. First, get the total weights from all of pixels in this mask if they have some weights, calculated by the above function.The total weight(accum) is accumulated and each weight(Wi) is stored. In the second step, the normalized weight(Wn) is obtained by the function Wn = (Wi/accum).
ReplyDelete