Threshole correction factor


Dear all,

How to determine the specific threshold correction factor value? Why the value is 0.8 in the module IdentifyPrimaryObjects in the example

Thank you for any advice!



The threshold correction factor value is usually advised to tune last , after you test all the different combination of threshold strategy and method .

For e.g.:

Global > Otsu > Two classes
Global > Otsu > Three classes (middle class assigned to Foreground/Background)

Adaptive > Otsu > etc. etc…

Only whence you really need to perfect the object segmentation, you would need to change the correction factor. Leaving it at 1 would most often be more robustly applicable to different sets of images (i.e. not to “overfit” to just one image)



Dear Minh,
Thank you very much for the reply!