Conference bulova::decw_jan-89_to_nov-90

I would guess that the 12 bits represent 4 bits each of R, G, and B, in X terms,
it is 12-bit TrueColor.  If that is the case, the only thing you can do on an
8-bit PseudoColor system such as a 8-plane PVAX is to choose a pallette of up
to 256 colors that are a best-efforts attempt to represent all of the different
RGB values in the original image.  Build a color map representing this pallette,
then translate the pixels in the image to the corresponding 8-bit pixels in the
color map.

This job is much easier on a 24-plane Firefox--all you have to do is turn the
12 bit TrueColor pixels into 24 bit TrueColor pixels by padding each of the R,
G, and B fields with 4 zero bits.

--PSW

    my understanding is that one can take 8-bit pixel data and display
    it on a bitonal workstation.  if so, how is this done, and does
    the method have applicability to the "12-bit to 8-bit" problem?
    
    thx.
    michael

The image services library can blue noise dither color images and get excellent
quality although it takes a while. I would suggest using ISL to color dither
from 4,4,4 to 3,3,2 and then saving the result as ddif for faster redisplay
of the images. I believe that the resultant images can be compressed in ddif
for disk space savings and faster network transmission. It is very hard to
differentiate a natural 24 plane image blue noise dithered to 8 planes from
the original. To try this put up a 24 plane natural image on a 24 plane firefox
use print screen set up to store color ddif to get the image and then display
the image on an 8 plane system with the cda viewer. the results are really
impressive. The only drawback is the time it takes to do the dithering.
The next generation VAX graphics hardware will be capable of doing the
dithering in hardware very quickly if we can get an x11 imaging extension that
supports that new hardware.
							Bill

	Since this is coming from a radiological device, I suspect the image is 
actually 12 bits of grayscale. There are two alternatives:

	o As has been suggested, you can use imaging services to dither the
	  12 bits of grayscale down to 8 bits which can be displayed. The
	  results will be quite good, but may not be acceptable to physicians.
	  The medical community (especially radiologists) have very strict
	  requirements about how image data may be processed for display.

	o Use a technique often employed in medical imaging applications known
	  windowing and leveling. This basically takes a slice an 8 bit slice of
	  the 12 bit space and redistributes it into 8 bit grayscale. Its a
	  combination of clipping and requantization.

	/John

>The image services library can blue noise dither color images and get excellent
>quality although it takes a while. I would suggest using ISL to color dither


(really a rathole)..  whats "blue noise dither"???

(I _think_ I know what dithering is... but why the "Blue")

"white noise" is random static.  The terms "pink noise" and "blue noise" skew
the random distribution in particular directions, the point being to give more
visually pleasing results than one would get with a purely random dither.  I
don't know the details of what constitutes "blue noise."

--PSW

    "Blue Noise Dithering" is an error diffusion dither technique which
    adds a random noise factor in to avoid correlated artifacts in the
    resulting image. The term blue noise is used to describe high frequency
    white noise, the power spectrum of which is zero to some high frequency
    and then flat thereafter.
    
    /John