I'm the person who asked the question Hamid linked to. Trust me when I say this, unless you are experienced with coding and/or ROS you will probably find this an incredibly painful route. Check my posting history, I wrestled for a long time with this. If I can save you the time and frustration, it will make me feel a bit better about the experience.
One for active image segmentation and
one for skin detection.
One thing I discovered by experience this year is that when PCL started to gain traction in 2011 a lot of the algorithms and programs for computer vision started to be developed over there. You can look at the tutorials on www.pointclouds.org and check the mailing list at www.pcl-users.org . Their support system isn't as good as ROS Answers but it's all you've got to work with. I also hear good things about OpenCV but haven't used it myself.
Although ROS and PCL messages can be passed between each other, the programming for anything related to computer vision has mostly migrated to OpenCV and PCL. ROS is going more in the direction of pure robotics with basic computer vision functionality. ROS Hydro even uses it's own version of PCL I believe.
For what it sounds like you want to do, check out PCL and OpenCV first. Unless you need Fuerte because you developed on it previously I wouldn't use it. It's deprecated, there's no support, it's a path of pain and futility.