On 25 May, Mark Vojkovich wrote: > On Tue, 25 May 2004, Charly Vidal wrote: > >> Hi, guys; I'm from BA, Argentine. >> >> I'm working on medical aplication. >> >> Has anyone any idea about how can I output and image >> with 1024 gray of shades? > > You can't. > >> >> I don't kown how to "create" a LUT with 10-bit . The >> type of file is DPX. > > Quadro has a 10 bit DAC in the sense that it can display > 8 bit per component data by passing it through a 256 element > lookup table that has 10 bit per component values in it. > That is, map 8 bits to 10. This is really only useful for > doing gamma correction. If you only had an 8 to 8 mapping > as soon as you changed the gamma from 1.0 you'd get duplicate > entries. Only NVIDIA's binary driver support the 10 significant > bits per map entry. This does not let you display more than > 256 greys. It merely lets you tune 256 greys to a finer > resolution. > Most medical applications actually display 256 shades of grey, using the deeper LUTs to adjust the brightness. It is very similar to gamma correction, but you are correcting the display to uniform linearized perceptual response. At last week's SCAR poster session there was an excellent poster on the perceptual differences between 8, 9, 10, and 11 bit display on LCDs. There was also a paper presented on the pixel level non-uniformity of LCDs and a demonstration of the perceptual difference between use of uniform per-panel LUTs and the more difficult per-pixel compensation. To get per-pixel compensation you need a manufacturing facility that measures the LCD panel at the pixel level. These discuss perceptual ability, not clinical impact. There is little (if any) clinical ROC or other studies to determine the clinical impact of these differences. It is current common practice to downconvert the 10 or 12-bit image data to 8-bits as part of window/level and image processing, and then display these using a perceptually linearized LUT. So it is appropriate to use the Quadro LUT in that way. That kind of display is considered medically appropriate for many purposes. R Horn _______________________________________________ XFree86 mailing list XFree86@xxxxxxxxxxx http://XFree86.Org/mailman/listinfo/xfree86
