How can i fix the threshold automatically from the histogram?

How can i fix the threshold automatically from the histogram?

Answer 1

[Walter Roberson]

First you need to define how you want to make the selection.

In the general case, a histogram of values could have complexity equivalent to a list of random non-negative integers, with up to half of the values potentially being peaks. There could be multiple peaks of the same height; there could be multiple major modes.

[MoonPie1]

Having three gray values, is there any function which thresholds automatically from histogram and displays it?

[Walter Roberson]

No, there is no function that combines all of those things.

You still have no defined how you want to make the selection; all you have done is restricted the number of bins without indicating how the threshold is to be chosen from the counts.

You might want to look at http://www.mathworks.com/help/images/ref/multithresh.html

But more likely what you are trying to work on is histogram equalization

[Image Analyst]

I don't think histogram equalization would be what she wants. I think that would just make the three gray levels 0, 128, and 255 instead of whatever they are. I think she needs to define what needs to be done after the threshold, for example call regionprops(), like

bw1 = grayImage == grayLevel1;
measurements1 = regionprops(logical(bw1), 'All');
bw2 = grayImage == grayLevel2;
measurements2 = regionprops(logical(bw2), 'All');
bw3 = grayImage == grayLevel3;
measurements3 = regionprops(logical(bw3), 'All');

If it's really NOT 3 gray levels but three humps in the histogram, then she'll have to come up with a automatic thresholding scheme to pick them out. Perhaps multithresh() would do a good job there.

Such speculation can often be reduced if the poster would have just included the image in the first posting.