How does the 'Image Segmentation Using the Color Thesholder App' determine the thresholds ?

Using the color thresholder app you can select regions you want to segment (be able to separate from the background). How does the color thresholder estimate those color thresholds ? The documentation (<http://nl.mathworks.com/help/images/image-segmentation-using-the-color-thesholder-app.html#buj7jf_-1>) doesn't explain the method used. Is each threshold the mean value of the channels of the selected regions ? Or does it use Otsu's method ?

Answer 1

[Image Analyst]

It doesn't decide on the threshold, you do.

You're supposed to grab the blue bars and drag them to a threshold value manually/interactively.

[nim3003]

My question isn't about selecting the thresholds manually by dragging the bars. I meant if I select a region (with the desired feature) using the app. Does the thresholder take the mean of the values of the different channels of the selected pixels ( in the region I selected) to calculate the thresholds for the entire image ?

[Image Analyst]

No. Again, thresholds are not calculated.

Not calculated from the mean and not calculated from anything else. Not calculated period.

Use the "Export" button and you'll see they're simply assigned with the values you specified.

[Image Analyst]

If you have it make a first guess at what the thresholds should be by clicking the Select Colors button to freehand draw a region then click Find Thresholds, I would think that it uses the same algorithm as in imseggeodesic() though I'm not sure because it does not say in the help. You can call them and ask them. The method for finding colors in imseggeodesic gives a paper that you can order and read to learn the algorithm.

[1] A. Protiere and G. Sapiro, Interactive Image Segmentation via Adaptive Weighted Distances, IEEE Transactions on Image Processing, Volume 16, Issue 4, 2007.