We currently have our image processing set up this way. As Alan said, the Driver Station is already getting the image, so we use that one. Then we send the relevant numbers (x position of top target, distance) to the robot over one of the available UDP ports as described in
2.2.9. UDP 1130 to be specific.
The camera has an http server hosting the image, and the driver station and the cRio both request an image from this server. Or perhaps it's just the cRio, and then the cRio forwards it to the DS; I haven't looked at the code in a while.