This question is more directed to those who are familiar with the 'feature' function developed by Crocker and Grier. The code is located at: http://www.physics.emory..../idl/kit/feature.pro I've been trying for some time now to develop a similar code that can be used for particles with a large size distribution. I haven't been able to find a way to modify the feature code to work for such a system, and I'm hoping someone might have a suggestion. The code uses a parabolic mask to convolve the image and find features, but it requires that the user specify the mask size, which assumes uniform size distribution. I've written a piece of code that uses an entirely different method (involving numerical derivatives) but it is much less efficient than the feature code, since feature makes use of IDLs speed with array operations. Another small issue with writing this code is that the images must be cleaned up somehow, which is normally done with a bandpass filter. Unfortunately, a bandpass filter also requires that an estimate of feature size be made, so I need to figure out another method for cleaning my images. If anyone has any ideas as to how I could implement such a piece of code, I would greatly appreciate the advice!