Fix image encodings to comply with ROS

[svepe]

In order to use standard ROS image tools with the published images they should have the correct ROS encoding (not OpenCV). For the full list of definitions see here.

[kohrt]

Hi,
thanks for the PR. I don't know why there are different types for uint16 in ROS, if there where just different names for the same type, I wouldn't care.
But changing to MONO16 isn't a good idea. Rviz works fine the way it is now, and for example image_proc checks for the CV types, see here.
The correct solution would be, to change sensor_msgs, so that both names result in the same string, for example const std::string MONO16="16UC1"; const std::string TYPE_16UC1="16UC1";.

[svepe]

The depth_image_proc checks for the OpenCV types which is fine for depth images, however the color images and the mono images should be with ROS types. Important tools such as cv_bridge assume that ROS types are used for a sensor_msgs::Image (see here).

[kohrt]

But it also checks the other ones, if you look further down.
Where does it result in problems, or which packages don't work, if the encoding stay like it is now?

[svepe]

I did not see those. The reason I did the pull request is that something from the package calibration_launch complains that it requires mono8.

[kohrt]

Here are the CV_TYPE checks: https://github.com/ros-perception/vision_opencv/blob/indigo/cv_bridge/src/cv_bridge.cpp#L83-L100

[mpatalberta]

How do you plan to handle intensity and rgb. The intensity IS NOT the color but is the strength of the return. PCL DOES NOT allow both intensity and rgb.

[kohrt]

@mpatalberta what do you mean? You can have multiple channels stored in a point.