Gaussian fitting of multiple spots

Hello everybody,

I have an image which is composed of a black background and six white spots in double precision.

I need to fit each of the six spots with a Gaussian profile. What's the best approach to this problem?

Thanks a lot!

Answer 1

[Image Analyst]

Do you need help with the finding, or with the modeling/fitting? Or both? If you need help with finding, you can threshold and use bwlabel() and regionprops() if you have the Image Processing Toolbox.

[Ed]

Thanks for your answer! I actually need help with both, but especially the modeling/fitting. I have the Image Processing tool and I will try the functions you suggested! Do you know what the best options for modeling and fitting are? Thanks again!

[Image Analyst]

See my File Exchange for how to threshold and find the centroids. http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

What you are asking for is essentially the "CLEAN" algorithm used by astronmers http://www.cv.nrao.edu/~abridle/deconvol/deconvol.html. They find the brightest point, fit a Gaussian to it and log that Gaussian to an output image. Then they subtract that model Gaussian image from the original image and repeat until all the spots are gone or dimmer than you care about.