computer vision - Tips for background subtraction in the face of noise -
background subtraction important primitive in computer vision. i'm looking @ different methods have been developed, , i've begun thinking how perform background subtraction in face of random, salt , pepper noise.
in system such microsoft kinect, infrared camera give off random noise pretty consistently. if trying background subtract depth view, how can avoid issue random noise while reliably subtracting background?
as said, noise , other unsteady parts of background might give problems in segmentation, mean lighting changes or other moving stuff in background.
but if you're working on indoor-project shouldn't big of issue, except of course noise thing.
besides substracing background image segment objects in try subtract 2 (or in methods three) following frames each other. if camera steady should leave parts have changed, objects have moved. easy method detecting moving objects.
but in operations might use have noise described. easiest way rid of using median filter or morpholocigal operators (opening) on segmented binary image. should remove small parts , leave nice big blobs of objects.
hope helps...
Comments
Post a Comment