View Full Version : Sub-pixel accuracy

11th November 1999, 04:49
What does "Sub-pixel accuracy" mean?
Someone? Please?

Intel Celeron 433, G200 8Mb AGP, Intel Atlanta 440LX, Soundblaster PCI 128, 96 Megs SDram, 4.3 WD Caviar HD, AOC 5VLR monitor 15". Besides that, a life.

11th November 1999, 18:05
it's probably that space between the pixels. If you look close (spit on the monitor first), you can see it

11th November 1999, 19:40

This is only a guess, but I think it means that Direct3D will render an image or texture using more pixels than are really available on the area of your screen where the image will appear. Then the final image is "averaged" to use the actual number of pixels. Since the internal calculation assumes a higher number of pixels than are really there, you don't lose as much information.

I guess that's a lousy explanation, but it's similar to Matrox using 32-bit color internally and then converting it to 16-bit for the final image if you are in 16-bit mode. Or maybe it's similar to dithering on your printer. Or maybe all of the above.

Perhaps the best example is this. You want to multiply 10.4 by 10.4 and you want the result rounded to the nearest whole number. If you do the internal multiplication at the highest accuracy and round the result, you get 108. If you round each number first and then multiply, you get 100. The true answer is 108.16, so obviously the 108 is more accurate.

Boy, that was long-winded. Did it make any sense??

Mark Veneziano

22nd August 2000, 14:37
When you draw a triangle to sub-pixel accurate co-ordinates, the actual, whole pixels drawn are affected slightly, especially with anti-aliasing on the edges