[vtkusers] Direct RayCasting on RGB(A) with VTK possible?

[David E. Jones]

Lisa S. Avila wrote:

> Hello Burkhardt,
>
> No, VTK does not directly volume render RGBA data. It is on the "to do"
> list, but will not be part of VTK 4.0.
>
> Lisa
>
> At 10:54 AM 11/21/2001, burkhardt.dorn at berlin.de wrote:
> >Hi all,
> >I would like to know whether VTK implements specific or
> >common ray-casting algorithms also on 3-dimensional
> >RGB(A)-images. The samples I inspected work all on
> >pure intensity maps and the color comes from a LUT.
> >Of course it is no problem, to convert RGB to Gray.
> >
> >But one might also try to perform a ray cast obtaining
> >the intensity of a given voxel by averaging the RGB of
> >its scalar and the color by taking it directly rather
> >than from a fictive LUT.
> >
> >Is this or something similar possible with VTK? A small
> >cxx or tcl sample would be useful... :-)
> >
> >Regards
> >Burkhardt Dorn
> >mailto:burkhardt.dorn at berlin.de
> >
> >--
> >berlin.de - meine stadt im netz. Jetzt eigene eMail-adresse @berlin.de
> >sichern!
> >http://www.berlin.de/home/MeineStadt/Anmeldung
> >

 I've  been dealing with a similar issue but  I think the solution must be domain specific so
that having a RGBA capability may not
 solve your problem.   Where does your data come from ?

My color data comes from a Laser Scan Confocal Microscope.  Different
materials are made to flouresce and the data is collected with a colored
filter on the camera.  The data for different materials is put into the
Red, Blue. or Green Channels but the color isn't real. Also there can
be more than three channels.

I would really like to have a reference which explains image processing
for the confocal microscope.  If any VTK users have information on this
I'd like to find out about it.   Thanks in advance.

If the colors represent materials there is method of putting the matterials
(i.e. colors) into a volume which is sketched out in a 1988 Siggraph paper
by Drebin, Carpenter, and Hanrahan.   Here's what I'm doing.  It's
pretty simple.

   Let Int(m) be the intensity for material m, m = 1,N       ( Int(m) is arbitrary )
   Let Opa(m) be the opacity for material m, m = 1,N      ( Opa(m) is arbitrary )

   Let   P(m)(i,j,k) be the probability that voxel(i,j,k) contains material m.  (We don't know
this yet)

         (i)  P(m)(i,j,k) is a good approximation for the volume fraction
         (ii) P(m)(i,j,k) can calculated using a  "probabilistic estimator" such as
               a "maximum likelyhood estimator".  I think this isn't too hard to write
               and there are simpler estimators.

    Therefore,    Intensity(i,j,k) = SUM (P(m)(i,j,k) * Int(m) ), m = 1,N
                         Opacity(i,j,k)  = SUM (P(m)(i,j,k) * Opa(m) ), m = 1,N

                       (Int(m) probably should be normalized by dividing by its range.)

    Maybe P(m)(i,j,k) = 1 / m would work in many cases.  I think that is equivalent to
    calculating V in the HSV color system.

    One simple approach bases P(m)(i,j,k) on the distance of the value of the
     intensity in channel m from the mean of the intensities in channel m.
     All this stuff is explained in Image Processing books like Gonzalez and Woods.

I MAY BE  GOING OFF IN A COMPLETELY WRONG DIRECTION FOR
CONFOCAL MICROSCOPES.  IF ANYBODY KNOWS A MORE
CONVENTIONAL APPROACH, PLEASE LET ME KNOW.

Thanks in advance,
Dave

--
David E Jones
Du Pont Central Research
Experimental Station, Bldg 320
Wilmington, DE 19880-0320


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20011126/44f45bcd/attachment.htm>