It mostly depends on the kind of noise you are dealing with. The first general remark is that getting ride of the noise during the acquisition is better than trying to filter it later.
So:
1- Try improving the quality of the images (lighting, camera settings, etc.).
2- Do image averaging (see developer exchange library or LabView examples).
3- If 1-2 are not working, you need to characterize the noise and the signal of interest to determine what kind of numerical processing might help.
