Detection of objects

My project is "Comets Recognition From Satellite Images".Initially I implemented subtraction and guassian to remove noise and blur from the images.I with the help of you found the objects and their intensities along with their pixel co-ordinates but there are many objects in the order of 90,000.I think there is some noise and need your help in removing noise.

[Daniel Armyr]

If you allready used the gaussian to remove blur, then you are pretty far along. If that was not enough, then I think you will ahve to provide a sample image to show how bad the situation is, and what the input data looks like.

Answer 1

[Image Analyst]

You need to register and subtract the images. I think that's how it's usually done. Alignment can be done with the Groth algorithm, that's the classic astronomy alignment algorithm but there may be more advanced and better versions now. Once aligned the stars are fixed and will subtract away and only objects that moved between exposures will remain (such as comets). Looking at a single frame is not good because often bright things, such as gaseous clouds, will have asymmetrical shapes. So you can't just look at a single frame and find comets by looking for non-spherical objects or fuzzy objects, though if you have a school homework project it could work with single images as long as they supplied you with obvious, super bright comets in the image that are guaranteed to not have clouds in them.