I am thinking that maybe adding color would allow you to condense a lot more information into a QR code. It would be good if we can have entire programs and media reconstructed just from scanning QRs.
For those asking about RGB, there are a few reasons why QR codes work better for now:
Black and white still reads fine in bad lighting or on cheap cameras. Phones are very optimized for scanning QRs so it’s more reliable and less error prone.
Video compression, colors shifting with white balance, screen type, all of those make it harder to tell colors apart so you need more aggressive error correction. The real gain ends up smaller than it sounds.
You can already transfer entire (small) programs and media with blastQR btw! Feel free to try and report back. The limit is around 4 GB files but that would take ages.
I feel like better calibration patterns could largely address those issues. Since it’s temporal you could even have a frame that just shows every single color combination, and then have maybe something in the timing pattern for example to correct for deviations after that. Compression might be a large issue though, since chroma is always compressed more than luminance. I feel like multiple bits per pixel might even be possible, depending on how good your camera is.
Comments
I am thinking that maybe adding color would allow you to condense a lot more information into a QR code. It would be good if we can have entire programs and media reconstructed just from scanning QRs.
For those asking about RGB, there are a few reasons why QR codes work better for now:
You can already transfer entire (small) programs and media with blastQR btw! Feel free to try and report back. The limit is around 4 GB files but that would take ages.
That said, there is a project called libcimbar that is improving and does work with colors: https://github.com/sz3/libcimbar
I feel like better calibration patterns could largely address those issues. Since it’s temporal you could even have a frame that just shows every single color combination, and then have maybe something in the timing pattern for example to correct for deviations after that. Compression might be a large issue though, since chroma is always compressed more than luminance. I feel like multiple bits per pixel might even be possible, depending on how good your camera is.
I think you’re right and it might be worth experimenting. For now I’ve created a wiki page to list techniques and research and write down ideas for performance improvements: https://codeberg.org/blast/blastqr/wiki/Performance-improvement-ideas
https://hackertube.org/watch/ExwqNreocpg
Sign in to comment. Sign in