How can empirical cumulative distribution function be helpful?

[DannieSheng]

On top of today's discussion, I made some tests for another grayscale image where the plant is brighter than the background.
The grayscale image is the saturation channel of the original RGB image.

The histogram:

The ecdf plot:

For this image, a threshold of 120 got me the binary mask like this:

My understanding is that ecdf is useful to visualize the distribution of object sizes, but not necessary useful for pixel intensity.

[nfahlgren]

Thanks for looking into this more @DannieSheng. I agree with you, it's not looking like eCDFs are helping in terms of picking thresholds. Maybe we should reorganize the PR to just add the object size function as a function in the visualization subpackage (not a sub-sub package). We could call it obj_size_ecdf?

[DannieSheng]

Sure, I agree.