Quote:
Originally Posted by vampmiller
It seems some people are arguing whether it exists or not even when others have taken photos to prove it does. And it bothers me people say photos show something that isn't there. If you can photograph it, it exists. If you can't see it when you use it, that's fantastic and I am totally envious.
|
Say rather, if you can photograph it, it is an effect which is emphasised by a camera's CCD sensor. I am now going to go off on a massive diversion of topic because this stuff is just really cool and if I babble about it enough others might come to agree, or at least might come to see why comparing photos of e-ink screens to see what they'll look like on a real device is a complete waste of time, and getting angry about it is even more of a waste of time.
A CCD sensor is not an eye. It is also not a magical detection device of things that 'really exist'. What 'really exists' is light in motion, oh wait no, an electromagnetic field with specific values at every point in space, oh wait no, it's a state vector describing a vast quantum superposition of fields and particles (we call it 'the visible universe') atop a relativistic spacetime metric... there is no way to detect the portion of this superposition hitting a surface with complete accuracy, even in principle and with utterly perfect detectors (Heisenberg's uncertainty principle forbids, as does monogamy of entanglement). "Reality" is, to a strictly limited degree defined by quantum physics, ineffable.
Now a camera CCD detector has a bunch of bad cells in it, and only three types of sensor, which respond to only three wavelengths of light (roughly: there is variation, so the same type of sensor in distinct spots will respond to slightly different wavelengths); a visible scene is reverse-engineered from the output of this with a whole bunch of manufacturer-specific corrections to try to figure out what colours and intensities to present to the human. This is clearly not 'reality' in any sense. Camera sensors are not reality detectors. They are not-terribly-good sensors of a few specific wavelengths of light. It is very easy for them to photograph things that are not there, as anyone who's ever tried to clean up the horrible mess that is a raw photo (straight from the CCD) without software assistance will know well! Getting the colours right is more or less an exercise in guesswork to start with, and the intensity often seems to waver madly across the image until you teach your conversion software the varying intensity response of different parts of your allegedly-but-not-actually-all-that-constant CCD sensor.
Equally, the human eye has no pretensions to observing absolute truth: it's an even more terrible squidgy fault-filled optical instrument with hopeless resolution outside one tiny area, with lots of blood vessels and neurons layered on top of it (!) which has subsequently been massively optimized by evolution (some of those neurons, being translucent, do double duty as light guides, for instance!) and given a huge pile of immensely powerful dedicated neural backing to get the mental model we call a visible scene out of the resulting smeary mess, via a bunch of inverse optics which derives a 3D scene from the 2D image on the retina: this is necessarily an approximation, and as optical illusions show it can reach wrong conclusions.
Some of the work the visual cortex does involves making really quite radical adjustments to perceived colour and contrast levels: these adjustments are optimized for making sense of objects in the real world, not for observing the levels of colour and intensity across a piece of paper or e-ink screen or for observing what is "really there". For instance, one adjustment made very early in the visual cortex is to use both sharp and gradual colour and intensity changes in an apparently continuous object as a clue to the 3D shape of an object with constant colour that is partially in shadow; so a Kindle screen with a colour gradient might look to you to be a screen at constant colour that is very slightly bent: note that if this were a piece of paper that really *was* very slightly bent, the appearance might well be identical. (Again we see that what we are looking at is never reality: it is a model encoded in transient neural firing patterns, that's all. It is no more reality than what comes out of a CCD is reality.)
So these are both necessarily approximations very far removed from anything like objective truth, and they differ significantly from each other: indeed, it would be astonishing if they didn't, since we don't actually know precisely how human vision works yet, so it's a tall order to emulate it precisely in an artificial detector. A particularly radical example is their handling of intensity and contrast, but colour is quite differently handled, as well, with most CCDs being equally sensitive to each of the three colours they can detect, with a little post-facto adjustment in the camera firmware which takes it a bit closer to the very uneven frequency response of human vision, optimized as it is for the light of the sun; relatively insensitive to blue light, etc. (Again, if you've done raw work you've had to do this yourself.)
e-ink screens are optimized for the human-eye sort of terrible optical instrument, not the camera CCD sensor sort of terrible optical instrument. A mythical all-sensing being that could somehow see what was Really There would surely observe that the EM radiation emerging from an e-ink screen looked nothing like what emerges from paper at all -- but this is just as surely irrelevant. There are few mythical all-sensing beings among Amazon's customers!
(apologies to kobayashi, it looks like there's one.
)