Tool (crazy idea): Image compression

[j4lando]

Problem

Scientific missions in space are highly limited by the amount of data that they can downlink. It would be amazing if Husky Satellite 2 could downlink images of earth's horizon for future research and development of FOUND. However for the aforementioned reason this is not possible. The question is can we do better than current lossless image compression techniques?

Potential Solution

I think we might be able to. Compression algorithms assume that there is nothing known beforehand about the image, but this is not true in our case. We can simulate pretty accurately what Earth might look like from space given a specific camera. Using this prior knowledge we could send down only the differences between our simulation and the real image. Hopefully this would be smaller. If this actually works, it would be revolutionary to how we think about data in space.

[sathvikkurap]

The main limiting factor in this approach is the effectiveness of the simulation that generates the image to compare to. If other compression methods are more accurate then we don't really need to implement this.

Another thing I can think of, if the simulation happens on board then in order to reconstruct the data then we would also need the exact parameters used to create the simulation, otherwise we will get a completely inaccurate image.

This is almost the same approach implemented for missions to Mars' moons: https://www.mdpi.com/2072-4292/15/23/5500#:~:text=Particularly%2C%20JAXA%27s%20Martian%20Moons%20eXploration,with%20the%20limited%20downlink%20rate.

One of my concerns is the processing power required to run this simulation onboard. Although I suppose we could somehow preload simulation images onto the satellite (but again we may run into issues with memory)?

This approach has also been done leveraging spectral/spatial diffs:

https://ccsds.org/Pubs/123x0b2e2c3.pdf

The proposed approach is a great approach and has been done before, but it is maybe not as applicable to CubeSats. At least that's what I gathered after researching for a bit. I could also be completely wrong.

[MahirEmran]

Was looking into this for a bit and it could be feasible I think? Could chat tomorrow during meeting
Looking at these papers (in addition to the CCSDS one Sathvik posted):

• https://www.merl.com/publications/docs/TR2011-045.pdf
• https://ntrs.nasa.gov/api/citations/20050156924/downloads/20050156924.pdf
• https://indico.esa.int/event/130/contributions/712/attachments/773/949/DSCAL_CCSDS_ImageCompress_FPGA_SEFUW2016.pdf - maybe better idea on how CubeSat can apply?
• http://spacewire.esa.int/edp-page/papers/ippw7_paper_358_Cabral_2010-25-05b.pdf

Obivously will have to look more, and decent chance I'm completely wrong