3 ms·
Well the thing about digital data is that it's all represented in an uneven compressed way. Take your average piece of paper with text on it, it's essentially a
by moffkalast 4y ago
Well the thing about digital data is that it's all represented in an uneven compressed way. Take your average piece of paper with text on it, it's essentially a 2D array of white and black dots. Remove half of all black dots and you can still mostly see what the text was.
Same is true with raw digital representations. Take on the other hand a jpeg which is a bunch of coefficients for sine waves, where every tiny change ripples out to ruin the entire thing. Even the basic 3 byte RGB per pixel representation is already compressed with only 8 bits being used to save 256 values. Flip the MSB and your value is already completely screwed.
An equal encoding to real paper would be using 256 bits to represent each colour, summing up the bits to get integer values. That way you can flip 20 bits and you'll still easily tell what the colour was, without any extra processing.
This of course comes at an exponential increase in storage requirements, plus we can be smarter than just using brute force. Taking a compressed representation and then adding on a few layers of error correction would likely be far more reliable while preserving most of the spatial efficiency. Of course then you need the software for unpacking and correcting to survive as well...