< Chronological >      < Thread >

1. The software release process, in both binary and source formats, is not stable.  There are slight incompatibilities between versions.  Followup on changes in the development environment (Qt, compilers) is difficult.  You must know this already.
2. The UI for controlling transfer functions is under-functional.
1. The widget for 1D function is too constrained.  The control over smooth/step incline/decline (right mouse button) is quite limited.  Somehow, I feel that even with this simple tool I cannot cover the full range of "window-level" functions (with smooth ends).  The other option, dragging the curve with the left mouse button, is worse.  I never know what shape the curve will take as I drag it, and it typically creates sharp discontinuities in the transfer function.  I hope that you'll find better UI mechanisms for this function.
2. The widget for 2D function (I mean the whole thing, not just the shapes one draws on it) is exciting but mysterious.  I am attaching several screen snapshots.
   
   The "HeadCT" image shows a pretty nice segmentation, which I could get even nicer by fiddling more with the widget.  However, I have a hard time understanding the meaning of these shapes I draw, and of the histogram behind.  The explanation in the manual is not clear enough.  You are trying to lay out both the data (voxel values, I guess) and the gradient in the same space.  At least for CT, and especially clinical CT, the meaning of the data is fairly clear, as you may know.  But the meaning of the gradient is mostly vague.  It could be describing contact interfaces between different materials in the specimen.  But giving that a visual meaning is very difficult.  What I see in the widget are arc-shaped clusters, but I don't know what they describe.  By the way, if the clusters tend to have arc shapes, perhaps you should add an arc-shaped classifier instead of the rectangle/triangle or other convex polygons.
   
   The "Transfer2D" image shows a test phantom that was CT-scanned.  The data histogram (X axis) shows a concentration of the intensities in a pretty discrete set.  This is sensible, because the phantom is made of a small set of materials with mostly homogeneous absorption coefficients.  But this data is spread out in a very unclear way.  As I try to focus on each intensity band (Band1 to Band3) I can't find any logical relation between them and material properties.  They seem to cover basically the same material, albeit difference spatial subsets of it.  Other than this, I was not able to perform a reasonable segmentation of this scan in any way: 1D tfqn, 2D tfqn and isosurface.
3.  It would help a lot if I could define a transfer function analytically, without dependence of the UI widgets.  But this relates to my next item.
4. I wonder if there is a way to use iv3d or tuvok as a plugin inside other applications.  Specifically, I am looking to visualize data such as the HeadCT within the context of a different application.  Suppose that I find a transfer function adequate for the task of the application.  Still, I would like to show the visualization window inside my main application window, not in iv3d.  This looks doable, but as of now complicated.  I am studying the iv3d/tuvok code, but so far I did not find out what to do.  It appears that the Qt dependency should be maintained, which I can live with.  But what's next?
   
   Perhaps this is part of your development plans.  Any tips?

--

Ofri Sadowsky, PhD

Scientific Consulting

0 Givat Brenner,  Givat Brenner, Israel  60948

Tel: +972-8-9443944

Mob: +972-54-3113572