Vote count:
0
I am currently making an attempt at gesture recognition on a Beagleboard, using Simulink. I have skin segmentation working, but there are some objects that are being segmented along with the skin tones that are not skin. I know a way to help alleviate this issue is by first white balancing an image, but I've run into a problem here.
I am using the YCbCr color space for the skin detection algorithm, and gray world, as far as I've seen, only works in the RGB color space. I have tried implementing the algorithm with a color space converter block in between, and it both slows down computation significantly, and actually makes the image worse for some reason. Rather than switching between spaces, I was wonder if there was a way to use gray world, or some other white balancing algorithm, in the YCbCr space?
Aucun commentaire:
Enregistrer un commentaire