Trying to use the PR2 in Gazebo, the simulated Kinect gives different image encoding in regards to the depth-map related topics. The simulated Kinect uses 32FC1 instead of the normal 16UC1 which is used by the real Kinect.
Now suppotrs 32FC1 images. They are converted (internally) to 16UC1. This takes about 1 to 2ms per image on my i7. Passing 16UC1 is still slightly more efficient.
Trying to use the PR2 in Gazebo, the simulated Kinect gives different image encoding in regards to the depth-map related topics. The simulated Kinect uses 32FC1 instead of the normal 16UC1 which is used by the real Kinect.