mccool@csri.toronto.edu (Michael David McCool) (11/14/89)
With respect to shades of grey: 64 SHOULD be enough for monochrome, if you adapt the distribution of levels to the "gamma" curve of your monitor. This will probably be a nonlinear distribution. And yes, I have done this (not on the VGA; we do medical computer graphic reconstructions with big CAD/UNIX type systems). You might also want to adapt the colour map to the image, such as providing the most colours around the highest probability colour. 256 levels gives you plenty of brute force, which works, of course. Purists will probably complain that with their eye 1cm from the monitor they can see a difference; but I can't. If you add colour, your discrimination goes up. One trick used in reading things like CAT scans, which have very subtle grey scale, is to use a "thermal" scale, with white as the brightest and dark orange at the low end of the scale. This REALLY improves discrimination, because you change colour and brightness at the same time. So if your goal is reading subtle grey-scale images, try colour!
rick@hanauma.stanford.edu (Richard Ottolini) (11/15/89)
Someone asked are 64 gray levels enough? The answer depends on the spatial freqeuncies and correlation scales in the image data. Rapidly changing image data can get by on as little as 16 levels (e.g. geophysical seismic data) while low frequency changes may need 7 or 8 bits (e.g. Landsat images). This result has been known to perceptual psycologists a long time (e.g. Landsat images). A quick test is to blink between two quantization images and see if you detect a difference.