Suggestion: include zdepth lossless depth compression

[roelofvandijk]

To decrease the bandwidth of streaming depth data, it could be interesting to add a post-proccesing option to losslessly compress the depth (in a different manner than described in the colorization/compression whitepaper).

An example implementation with a permissive licence can be found here: https://github.com/catid/Zdepth

[MartyG-RealSense]

Hi @roelofvandijk A depth compression format called Z16H was added to the SDK in version 2.32.1, along with a processing block (rs2_create_huffman_depth_decompress_block) for using Huffman decompression to decompress the Z16H format.

https://github.com/IntelRealSense/librealsense/wiki/API-Changes#version-2321

https://intelrealsense.github.io/librealsense/doxygen/rs__processing_8h.html#ae2a413b273e6134503701e4fa51b3793

Though these functions are still documented in the SDK, I am not aware of any applications or scripting references that make use of the Z16H format, and in late 2020 the RealSense ROS wrapper added mechanisms to avoid using it due to it being 'currently unsupported'.

It may be a format that exists in the RealSense SDK (see the list of formats in the link below) but is not exposed to the end-user for use in applications

https://github.com/IntelRealSense/librealsense/blob/v2.32.1/include/librealsense2/h/rs_sensor.h#L88

[MartyG-RealSense]

Hi @roelofvandijk Do you require further assistance with this case, please? Thanks!

[roelofvandijk]

Hello @MartyG-RealSense, that is great news, thank you.
It would be nice to know whether this compression is available or just for internal use, I cannot find it in the docs/API/source code. If it is for external use, a benchmark would be nice.

[MartyG-RealSense]

I researched the issue further and located a C++ script shared by a RealSense user that uses Z16H as the depth format when recording to a bag file.

#8726

[roelofvandijk]

Hello @MartyG-RealSense, thank you, I will try that!