3 ms·
It‘s described in the README. The video has 2x2 pixel blocks that are either black or white, so each one signifies a bit. So a 1920x1080 frame encodes 518,400 b
by manmal 4y ago
It‘s described in the README. The video has 2x2 pixel blocks that are either black or white, so each one signifies a bit. So a 1920x1080 frame encodes 518,400 bit = 64.8KB
The assumption is that video compression won’t mess up those blocks beyond recognition, so you should retain the information as long as the rendered resolution and bitrate don’t drop too low.
Maybe this could be improved by e.g. using 32 colors instead of 2, and bumping the block size to 3x3 (for safety) which should yield ca 144KB per frame.
- LocalH 4y agoThe block size should honestly be tuned for the codec in use, chiefly to determine the best block size to fit with the codec's macroblock size. That's usually either 8x8, or with newer codecs 16x16. I feel like something like maybe 8x2 would be smart, and I like the idea of monochrome for resiliency, since chroma is downsampled. The fewer possible pixel combinations you have within a macroblock, the better the compression will probably end up being as well. And 8x2 would somewhat evoke the look of the old video backup systems as well, for the fun of the nostalgia of that.