Slashdot Mirror
Human Eye's Oscillation Rate Determines Smooth Frame Rate
jones_supa writes: It should be safe to conclude that humans can see frame rates greater than 24 fps. The next question is: why do movies at 48 fps look "video-y," and why do movies at 24 fps look "dreamy" and "cinematic." Why are games more realistic at 60 fps than 30 fps? Simon Cooke from Microsoft (Xbox) Advanced Technology Group has an interesting theory to explain this all. Your eyes oscillate a tiny amount, ranging from 70 to 103 Hz (on average 83.68 Hz). So here's the hypothesis: The ocular microtremors wiggle the retina, allowing it to sample at approximately 2x the resolution of the sensors. Showing someone pictures that vary at less than half the rate of the oscillation means we're no longer receiving a signal that changes fast enough to allow the supersampling operation to happen. So we're throwing away a lot of perceived-motion data, and a lot of detail as well. Some of the detail can be restored with temporal antialiasing and simulating real noise, but ideally Cooke suggests going with a high enough frame rate (over 43 fps) and if possible, a high resolution.
the early transparent calcite crystal lenses on the eyes of trilobites and even older compound eyes of some ancient arthropods are proof your religious fantasies are complete bullshit. The eye evolved, and we even know the creatures on the paths.
Hallucinations of the dying don't constitute proof of anything. Your god is merely a group psychosis constructed by goat and sheep herders less than three thousand years ago.
The other reason is the "sensors" we have are quite poor - the eyeball itself is actually a very low resolution device - the high resolution center part of the eye covers such a narrow field of view that it's practically useless if it was a fixed camera, while the peripheral vision is so low res it's unusable.
Instead, what happens is we evolved a gigantic amount of wetware to process the image into a high-resolution image we perceive - the brain does a lot of visual processing, and the eyes rapidly move (or oscillate) to move the sharp high-res center vision around to give you a much higher "virtual resolution" than the actual Mk. 1 Eyeball can achieve.
Of course, this visual processing comes at a price - optical illusions abound because it's very easy to trick the wetware into seeing things that aren't there, because the information is often interpolated, shifted in time, etc.
Such an imaginary construct does indeed fit the definition of psychosis by DSM, and in other cultures "near death experiences" can take vastly different forms. Hence, your links are to worthless imaginings. It is past time for society to admit religion is an irrational, often dangerous delusion that has done far more harm to people than good. It enables war, abuse, theft, and mental illness and is despicable remnant of the worst of human's history.
That's like saying laserdisc is digital "because it's got pits and non-pits". Except that the length of the pits and non-pits is very much analog. (It's a full-bandwidth FM signal driven to maximum overmodulation. VHS does a similar thing.) In other words, the digital-ness becomes analog if you look even closer.
#naabhaprzrag, #sverubfr-000, #agi-fcbafberq, negvpyr[pynff*=' negvpyr-ary-'] { qvfcynl: abar !vzcbegnag; }
Current motion blur techniques are smarter than you seem to imply, but we're still really far from anything resembling movies. Most implementations use per-pixel blurring (not blocks of a certain size), sometimes across multiple velocities per pixel, but they're still gross approximations which lack information (such as what's behind the frontmost objects). Re-rendering the scene multiple times at different time points, on top of being extremely time-consuming, looks really bad. You need something like 50+ images per frame to create the illusion of smoothness, and at that point you're better off simply presenting 100 frames per second and letting the human eye apply blur.
Offline techniques use stochastic sampling instead, so you're still effectively re-rendering the image a certain number of times at different time positions, but the difference is that each pixel uses its own set of unique time positions, which trades banding (i.e. being able to clearly see the different time steps) for noise. This sort of technique can be applied in real-time using algorithms like stochastic rasterization, but we're quite far from seeing that used in actual games.