|
|
Dear coders and ImageJ specialists,
Sorry if my question is very naïve, but I'm really struggling right now.
I'd like to know what is the best auto-theshold determination algorithm
to segment vesicules in fluorescence images (that is, bright spots on a
dark background, but with variation in spots/background intensities
between two images).
I don't want to go into watershed, because I don't need to individualize
particles, just detect what's above background and what isn't. A simple
condition like "5% of the image pixels" will not work because sometimes
there are very few vesicles, sometimes there are a lot.
The built-in "Autothreshold" in ImageJ is a bit too variable : some
times I get just half of the vesicles, sometimes it takes a lot of the
background. I started trying to do something with edge detection, but
difficult to generate nice, round, closed ROIs from the edge detection
image.
Do you know how I can do that ? Or alternatively, do anyone know a link
to a good resource on different auto-thresholding algorithm, how they
work and what to use them for ?
Thanks a lot !
On a related note, is it a known fonction that directly outputs the
integrated density of all pixels above a certain threshold I(t) as a
function of the threshold value (t)? Something like an histogram,
multiplied by the value for each value (0-255), and then summed from t
to 255, for each value of t between 1 and 255 ?
And is there a common name for the histogram multiplied by the intensity
value (histogram x 0 to 255) ?
--
Christophe Leterrier
Postdoc
INSERM UMR641 Neurobiology of ionic channels
IFR Jean Roche - Mediterranee University
Marseille, France
|
Locked
