How to exclude NaN values from evaluation on raster data

I have a raster which have NaN values. I would like to evaluate graythresh for my data, but the NaN values are taken into account and as a results I have 0 (zero). Is there a way to exclude NaN values from evaluation or remove them from data?

Answer 1

[Sean de Wolski]

x_no_nans = x(~isnan(x))

[Walter Roberson]

... and then graythresh(x_no_nans). graythresh() does not care what order the data is in.

[Karolina]

I have one more question, if the implementation of Otsu method in matlab is sensitive for very small values? I have one dataset with values between 0-1, but the majority of values is smaller than 0.004, so evaluated threshold is 0. The threshold is visible on image, but the algorithm cannot find it, and I think it is due to very small values.

[Sean de Wolski]

What is EM, the effectiveness metric output? Otsu is not great for all images, it's really only for when there is an object in the foreground.

[Karolina]

I have two datasets where the threshold is zero. EM is for these data: 0.9862, and 0.5902