I absolutely grant that discretized images are dominantly low resolution approximations. This does not mean, nor does it imply, that “pixels aren't little squa

I absolutely grant that discretized images are dominantly low resolution approximations. This does not mean, nor does it imply, that “pixels aren't little squares; they're sample points.” It was that claim I was objecting to. The media onto which you are discretizing, and the method you use to do so, can be squares, without at all contradicting the bulk of your most recent comment. You could discretize onto emoji, using the colours, gradients, and shapes that best fit the ground truth image, and it would still be true that this is a low resolution approximation, and yet the pixel—lit: the picture element, the unit of display—is for sure more an emoji than a zero-size sample point! (Addendum: to be clear, I was objecting to ‘sample point’/‘point sample’, cf. “other than the exact center-point of each grid square”. I don't object to the claim that these can be considered samples at all, since that is a very general term.)