Peak detection algorithm

Guillaume Gay guillaume at
Fri Nov 30 06:24:25 EST 2012

Hi all,
I have implemented the (2D) Gaussian peak detection method described in 
Segré et al. Nature Methods *5*, 8 (2008) 

In short it is a patch based detection where a likelihood ratio 
(Gaussian peak vs background noise) is computed on a moving window over 
an image. The originality (and interest for my kind of fluorescence 
microscopy problems) is that this is followed by a /deflation/ step, 
where detected peaks are subtracted from the original image and a new 
detection is performed on the so called deflated image. Also it is noise 
resistant and there are not so many knobs to adjust (only window size, 
sensibility and peak radius).

Is there an interest to include this in skimage?

The code is available on github there 
<>. I imagine there's room for 
improvements, and more compatibility with skimage standards.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the scikit-image mailing list