Get RGB image of Segmented Objects?

edit

Yep, just requires a few steps.

1. Choose which camera you want to use (on a PR2, narrow_stereo might be a good choice). I'll refer to this as camera.
2. Create a [image_geometry](http://www.ros.org/wiki/image_geometry).PinholeCameraModel using camera/camera_info.
3. Let's say you have a PointStamped called object_center which represents the center of mass of some detected object. Transform that point into camera's frame using a tf.TransformListener.
4. Project that point onto camera's image plane using your camera model's project3dToPixel.