hutch@fps.com (Jim Hutchison) (11/15/89)
In <391@ucsvc.ucs.unimelb.edu.au> U5569462@ucsvc.ucs.unimelb.edu (DAVID CLUNIE): [contemplation of 6 bit DAC input/output deleted] >Which brings me to the point that I have read somewhere recently (can't >remember where) that the human eye CAN'T distinguish any more than 64 different >shades of grey. Is this so ? Do people beleive it ? Nope, I don't believe it. Smoothly shaded surfaces will develop visible banding, shadows will have little definition, highlights will be boring. These are all the nifty things you will get with a limited color space. 64 is not all that bad though, much nicer than 32 or 16. Try comparing 64 to 256 shades on shaded spheres or high quality digitized flesh tones. I mention the need for high quality digitization because if you already lost the shades in the input you'll never get a chance to see them in the output. We could venture into biology. Since my experience has been with seeing and not sight, you will have to hit those books on your own. There seems no good reason to stop at 8 bits either, 10 may be appropriate. There probably isn't a monitor yet for projecting an average scene on a sunny day (near blinding sunlight and dark shadows under rocky outcroppings). Just for kicks, good b&w film at 30-40 bits of grey. -- /* Jim Hutchison {dcdwest,ucbvax}!ucsd!celerity!hutch */ /* Disclaimer: I am not an official spokesman for FPS computing */
jbm@eos.UUCP (Jeffrey Mulligan) (11/16/89)
hutch@fps.com (Jim Hutchison) writes: >In <391@ucsvc.ucs.unimelb.edu.au> U5569462@ucsvc.ucs.unimelb.edu (DAVID CLUNIE): >[contemplation of 6 bit DAC input/output deleted] >>Which brings me to the point that I have read somewhere recently (can't >>remember where) that the human eye CAN'T distinguish any more than 64 different >>shades of grey. Is this so ? Do people beleive it ? >Nope, I don't believe it. Smoothly shaded surfaces will develop visible >banding, shadows will have little definition, highlights will be boring. >These are all the nifty things you will get with a limited color space. >64 is not all that bad though, much nicer than 32 or 16. The question as posed is not quite what hutch@fps answered. Imagine doing the following experiment: start with a black field; now increase the luminance of an adjacent field until the difference can just be seen. (This is called a just-noticeable difference or JND). Now, keep doing this. Eventually, the added fields will become so bright that the eye will adapt and some of the dark gray fields will look black. At this point count how many JND's span this range. A general rule of thumb is that the JND is a fixed fraction of the base level (Weber's law), usually around 1%. It is fallacious to infer the number of JND's which span the visual range from observations using graphic's systems with so-many-colors because the sampling of the luminance range by the device may not match the differential sensitivity of the visual system. On an 8 bit system, 6 may be discriminable from 7, while 216 may not be discriminable from 217. -- Jeff Mulligan (jbm@aurora.arc.nasa.gov) NASA/Ames Research Ctr., Mail Stop 239-3, Moffet Field CA, 94035 (415) 694-3745
rsingh1@dahlia.waterloo.edu (11/16/89)
So.. What was this all started by? 64 shades of grey enough? Ok. First, people talk about bits and things, but you have to understand that the human body is analog, and organic. Which means that in some situations, it is more responsive to certain things than at others. This makes it dificult to define limits of sombodies ability to do something or other. But a way to tackle the colour question would be to use an analogy like this: Weight ratio detection stuff. That is, if you have to weights, one in each hand, and they were both equal. Your brain might fool you into thinking one is heavier, or you might just say you can't really tell. Subsequently, if you increase the weight by one gram, and the initial weight was about 1kg, would you notice? Doubtful. But once one weight is x% heavier than the other, you notice. Likewise, if one colour is x% brighter than the other, you notice. Just like the stars in the sky at daytime. While they are still there, as well as the ambient lighting present at night, you can't see them. Our visual sensibilities have been thrown out of balance. Also you have to take into account that the eye has the rods, cones and other junk, with each component reacting to different stimuli. You also should know that the monitor doesn't re-produce the entire spectrum. It lacks in quite a few extremes, most notably in the greens with a bit in the violet and crimson. Compound this with the different sensetivity ranges of different intensities of the colour and you can see that it all gets very very complicated. In simple grey-scale, it's good to have plenty (256 shades, for me, is a decent working number) of greys available. Why? Well, because for images, you want a good range for the overall ambiance of the image. EX. If the image was a photograph, well ballanced, with a high-spread of greys ranging from bright to quite dark, then fewer greys could actualy do. But if the image is basicaly dark, then it would never use the brighter greys, so your palette of 64 greys now becomes a palette of around 10 greys. Here, the eye picks up lots of stair-casing. The effect is worse with colour displays. Consider graphics by an amiga, compared with a VGA. That's just the difference between 4 and 6 bit. (4096 and 262144 colours (or shades/tint's/tones/whatevermakesyouhappy), But pictures done on the VGA, even at the lower resolutions appear MUCH less grainy. Because the shadows and dark regions have more degrees that they can fade through than the amiga (dozens more). Where on the amiga, a shadow would be 2 shades (an obvious staircase), vga has a smoother one. It's just the simple fact that most people can tell the difference. A trained eye (like mine :-) can tell the # of pixels or resolution of scann's and things about as well as a printer can tell what one kind of offset type from another, one version of a font from another. And while most of us won't pick up on it consiously, we do 'see it'. And as time goes on, and you happen to be walking by a display where Itarbm computers are sold, you'll notice if the Itarbm computer is using only 16 bit graphics, and your workstation has 32. Just wait and see! -------------------------- Paul Sop - Prez #2 of Spaghetti Western Words and Images
thant@horus.sgi.com (Thant Tessman) (11/18/89)
In article <18219@watdragon.waterloo.edu>, rsingh1@dahlia.waterloo.edu writes: > So.. What was this all started by? 64 shades of grey enough? > > Ok. First, people talk about bits and things, but you have to understand > that the human body is analog, and organic. Which means that in some > situations, it is more responsive to certain things than at others. > Actually, the human brain isn't really analog or digital in the sense that humans like to think when they build machines. Inputs into a neuron are 'digitized' in the sense that the neuron either fires or it doesn't. (This seems to be a very important trick for maintaining signal quality.) The information transmitted is contained in the rate at which cells fire. In this sense the brain is (sort of) analog. I don't know if someone has come out and stated it outright yet, but the eye is not only not very good at telling the difference between absolute brightness levels, but it doesn't even seem like it tries to be. What the eye does very well is judge relative changes in brightness (and colors) and turn that information into edges and surfaces. This makes sense because the important information to gather from the environment is information independent of the lighting conditions. So if one block at a certain brightness is shown next to *but not touching* another block at a different brightness, the eye will be much less sensitive to the difference in brightnesses. If, however, they are touching, it is easy to see very small differences in brightness as a line running between them. I'm pretty sure (although I have no proof) that the people who make claims like "64 shades of grey are enough" are doing the first kind of test, not the second. It is possible to tell a difference between an image using 36 bits of color and the same image using 24 bits of color, let alone the difference between a 6 bit and an 8 bit monochrome image. thant -------------------- There are 336 dimples on the standard golf ball.