Optimization of parameters

[JulienPepin]

Hi everyone,

I'm currently doing a training period in a lab where we make thin layer deposition and I use SExtractor(SEP) to find, count and describe defects in depositions.

I was wondering how the filterthresh works in the background estimation?
I was wondering too what exactly is the filter_kernel? Is it the représentation of the shape of expected objects?

This image (400 x 400) was done with a bw & bh = 4, fw & fh = 5, fthresh = 0 for background estimation
and detection parameters :
threshold=3.5, err = bkg.rms(), minarea=1.

You can see a defect in the top-left corner (near 60 x60)

The probleme is in the same image set, those parameters are pretty good but in some cases they aren't enought to detect all defect. but if i choose a lower threshold for detection, I'll have false detections in other images.

I try to find a correlation between image properties (like median, mean or amplitude of value) and parameters i have to apply to have an optimized detection but I've to understand how those parameters work.

Thank you in advance for your help.
best regards.
Julien Pépin