Doodles for Research.
Next Goal: Shortening the Learning Procedure by Illustrations.
Friday, January 20, 2012
Better, when reversing the contrast-aware distribution
The reason for this is because the priority scheme is working for that. They are still worse than traditional hard thresholding. I am guessing if adding the adaptive threshold control will be much better than this.
Yes, it's interesting. I think some care needs to be used when selecting local thresholds -- the standard thing (average value in local neighbourhood) results in oversplitting near-flat regions. That's why I started looking at histogram shape, with the idea that bimodal distributions can be detected and used to decide the threshold. Not every distribution has a clear enough threshold obtained simply from the histogram, though.
Yes, it's interesting. I think some care needs to be used when selecting local thresholds -- the standard thing (average value in local neighbourhood) results in oversplitting near-flat regions. That's why I started looking at histogram shape, with the idea that bimodal distributions can be detected and used to decide the threshold. Not every distribution has a clear enough threshold obtained simply from the histogram, though.
ReplyDelete