[vtk-developers] A new way to look at images

[David Gobbi]

On Thu, Jan 27, 2011 at 8:09 AM, Dean Inglis <dean.inglis at camris.ca> wrote:
> Hi David
>
> thanks for the update, now I'll just git pull instead of downloading.
> Great that you got the clipping working on the reslice output!
> Looks like you are making that possible with a stencil to mask out the
> reslice background?

I considered that, but decided it would be easiest for reslice to just
set the alpha to zero outside the mask region.  There was a third
possibility, which would have been to render a polygon instead of
a quad.  But I chose not to do this, because using polygons would
have made compositing more difficult.  By using quads, I can just
select a quad that is large enough for all the images that will be
composited, and eventually I can multitexture all the images to the
same quad, and ideally make it possible to use a fragment program to
control the compositing.  (All of this is for the distant future... but
I want to make sure that my basic design makes it easy to accomplish).

> I also added a vtkOutlineFilter pipeline to my tcl
> example (attached) to show where the reslice output fits into the 3D data
> bounds.

Nice.  I'll take a look.

> Would it be relatively straightforward to write a mapper that incorporates
> vtkImageProjection to do oblique MIPs simulated x-ray projections etc. ?  It
> might be interesting to just
> have the plane normal adjustable with the camera focal point fixed at the 3D
> center of
> the image volume.

Yes.  I've planned this from the start.  There will also be the
ability to choose the thickness of the slab that will be projected.
Another thing that I'm thinking of is N-up views.  It's easy for reslice
to extract multiple slices, and then the mapper can reorder them into
a single texture that contains an N x M grid of images.

It will also be possible to control the plane normal independently of the
camera's view plane normal.

A lot of these things won't be available for a while, though.  My immediate
concerns are 1) a basic 2D resliced view, 2) image blending, 3) picking.
And after that I'll have to go through and clean up the code, which is a
mess.

 - David