Comment by londons_explore

There's an amazing way of encoding data called "gold codes". By having enough pixels like this, you can correlate the image with the expected pattern, and successfully extract data even though no individual pixel is visible.

It's used in GPS transmissions to allow decoding signals considerably weaker than the background noise. Because the receiver is aware what the signal should look like, it can extract it despite all the noise by averaging across all the samples.

It does require perfect alignment though, which might be tricky considering camera lens warping, etc.