Anyway to generate pointcloud from compressd depth image?

[ventusff]

For depth camera of relatively high resolution and high frame rate, or for computing platform which have limited resources, the pointcloud data is generated direcly from camera/depth/image_raw(or other raw depth images), which would cause the pointcloud data is relatively huge.
During my test with Astra Pro with rqt_topic (topic monitor), /camera/depth/points could reach 80M/s data rate. To generate this message consume a lot of CPU.

So I'm wondering if there is a possible way to configure processing.launch.xml, so that the package could generate pointcloud from topics like /camera/depth/image_raw/compressed?