pouët.net

what you suggested is practically dithering; the trend goes the opposite direction though: sub-pixel rasterization

what about extending the spectrum range .. so you can get a tan with your monitor.

I probably should patent that.

"Most computer displays have pixels made up of multiple
subpixels"

that's what i meant, actually...

using 1280x960 to rasterize (can it be called this way?) a
640x480 image with 3x 24bit subpixels per pixel ... either
ignoring the 4th pixel or using it for brightness or something...

really need to try this out tomorrow.

btw, anes ... it's called: Hue, Saturation and Value.

no, that's exactly what i didn't meant.

so ... what did you mean? xD

hvs

and what does hvs stand for?

human visual system

oh. k i get it.

so it's probably really a stupid idea after all.

somebody wouldn't perceive more colors but less "banding" ?

...go away?
Demos run at 60fps... who as time to notice banding? I don't. I don't even shoot my photos in RAW mode! (12 bits in my machine) I should.

doc: by alternating between two different colors rapidly, what you get is the average brightness between them. that gives you one more bit resolution, independent of how many levels you had originally.

so e.g. pageflipping between two 24bit images gives you 27bit net bit depth (1 bit gain per channel). doubling each line then using different color values each line (fakemode-style) gives you another bit per channel, which is 30.

anything beyond that quickly gets very very visible.

the big gains came from the fact that you used these tricks in *paletted* modes, where you had a large number of colors to pick from (2^18 for VGA) then palette entries. a palette with all available shades of red/green/blue easily fits, and with pageflipping (and order of r/g/b lines changed on the other pages, so it tended to average out), it looked somewhat okayish. allowing different colors on each page to get theoretical 21bit/pixel was a cute trick, but not really practical then, and certainly not now .)

xernobyl: "who as time to notice banding? I don't."
depends. with 24bit it's ok, but a lot of demos significantly lose precision during lighting or postprocessing. it annoys me no end that theta has some notable banding in places (hard to get rid of with ps1.x perpixel lighting), and i also think that protozoa (or animal attraction in some places, or tons of other demos) lose visual impact because of it.

Dumb question: would it be possible to calculate the luminance channel seperately at a higher precision? Would solve banding without having to perform high precision calculations on all three channels. (this is of course assuming you're using hardware accelleration)

not really - having to do this in calculations is pretty awkward: inside shaders there's no point, and for inbetween storage (i.e. between passes) there's no official format that has this kind of asymmetric allocation, so you'd be doing a lot of packing/unpacking work in shaders - which isn't much fun (and not particularly fast either). related ideas are used in some HDR formats for storage though (most notably RGBE).

another problem is that the banding artifacts aren't independent of color channel and intensity - e.g. they (apparently) tend to be relatively visible on images of dark blue skies, or so i've been told anyway (didn't bother enough to check). i haven't heard of similar problems with green or red.

This thread makes me want to kill myself in disgust

oh now i get it. thanks ryg :D

hdr displays, maybe http://www.cs.ubc.ca/~heidrich/Projects/HDRDisplay/

ryg: I've noticed banding in some demos but I don't really think that they lose "lose visual impact"... Outside demos banding only makes me "yuk" on skys. Sky is a very normal place to find banding in DVDs and photos! In demos I notice banding in A LOT of TBL demos and I don't find it loses visual impact. It's design. :D

Or use low res .gif with 50 colors.

Can someone show an example pic of this banding...