6
votes

I have created a perfect skin detector, but it detects ALL skin. What are my possible options to separate the hands from the rest of the skin?

I don't want to use haar identifiers, so please don't suggest it.

2
Did you mean "hair", or is "haar" some kind of algorithm?Marcelo Cantos
haar its a type of machine recognition algorithimuser1017485
How general do you need your detector to be? Will there be a (set of) particular gesture(s) to detect?Bart
I just need to detect the actual hands Im using convexity defects for gesture recognition I am having problems with isolating the hands from the rest of the image, not detecting gestures.user1017485

2 Answers

1
votes

pretty sure that SIFT would work for this kind of thing, its much more powerful than haar.

take a look at this paper: http://www.google.com/url?sa=t&rct=j&q=sift%20hand%20recognition&source=web&cd=1&ved=0CBsQFjAA&url=http%3A%2F%2Fciteseerx.ist.psu.edu%2Fviewdoc%2Fdownload%3Fdoi%3D10.1.1.103.4092%26rep%3Drep1%26type%3Dpdf&ei=sru5TtmSHurciAK6iJnyBA&usg=AFQjCNEsGRxKVLR1xIGSRT1XAQ2WPeehww

seems fairly straightforward

I'm been doing some SIFT as well, and my project plans does involve this type of thing, just that I haven't got to it. If you make some good progress, please be sure to fill me in.

2
votes

My suggestion is to first segment out the different skin regions through blob detection.

You can then eliminate the unwanted blobs by area and shape thresholding A bit of filtering before these steps would help a lot