Does depth_image_proc/point_cloud_xyzrgb take distortion parameters in camera_info into account - or do I have to undistort the depth image myself prior?

edit

Howdy folks

Using the zed_wrapper, I have logged (rosbag) rgb and depth images from a ZED stereo camera, along with camera_info topics, but omitting the actual point cloud data in order to conserve drive space.

When I create the point clouds offline using the the depth_image_proc/point_cloud_xyzrgb nodelet I notice that the geometry is warped.

Question: Do I have to un-distort the rgb and depth image first (e.g. using image_proc) before passing it to depth_image_proc/point_cloud_xyzrgb or does the latter take the distortion parameters automatically into account when creating the point clouds?

Thanks

Galto