Today I’ve learned about QOI (Quite OK Image) format (while reading GIMP 3.0 release notes!). I followed the urge and implemented image_transport for it.
It can be used for lossless compression of RGB or RGBA images. Its compression ratio is comparable to lossless PNG, but encoding is 10-50x faster and decoding 2-4x faster. Check the benchmark.
Thanks to our package image_transport_codecs, qoi_image_transport offers not only the normal image_transport functionality (on subtopic “qoi”), but it also provides a pure C++/Python interface to do image encoding and decoding without the need of going through the tedious and unreliable pub/sub cycle. If you’re interested, I’m curious about your opinions on image_transport_codecs and whether such thing should be used as the general backend for image_transport.
Slightly related advertisement:
If you now want to convert some of your bag files and images in them to QOI format, you can utilize our cras_bag_tools/filter_bag utility, e.g. with a recompress config. But feel free to explore also the other implemented filters, e.g. for topic filtering by name/type/size, TF filtering, remapping of topics or fixing wrong frame_ids .
Even less relevant advertisement:
The package is fully REUSE-compatible. This means automated license checking and composition of SBOM is possible. Join the movement and upgrade your packages with REUSE! There is even a handy Github Action for verifying compatibility.
