Conveying Structure
From Visualization Sp06
Lecture on Mar 9, 2006
Readings
- Smart Visibility in Visualization. Viola and Gröller (html)
- Using Deformation for Volumetric Browsing. McGuffin et al. (pdf)
- Interactive Image-Based Exploded View Diagrams. Li et al. (html)
Optional Readings
- Non-Invasive Interactive Visualization of Dynamic Architectural Environments. Niederauer et al. (html)
- Cutaways and ghosting: satisfying visibility constraints in dynamic 3D illustrations. Feiner and Seligmann (Springer)
- Volumetric Illustration: Designing 3D Models with Internal Textures. Owada et al. (html)
Cynbot - Mar 06, 2006 09:54:45 pm
In "Using Deformation for Volumetric Browsing. McGuffin et al." the authors cite my work from 2002 [1]. I think a comparison to our software is a strange and un-necessary motivation for their work.
Nchentan - Mar 09, 2006 02:00:58 am
Comments for the reading "Using Deformations for Browsing Volumetric Data": The tools presented seem to be very useful. The papers also nicely summarize various techniques for peeling layers and associate the 3D versions of each, as well as provide hints on the implementation. It does not seem obvious, however, to efficiently implement the technique on data that has connectivity. Another comment on the examples, I find it rather disturbing to use peeled human head as the only graphics in the paper!
AaronHoover - Mar 09, 2006 12:24:20 pm
These readings reminded me of a visualization tool called JIV for visualizing structure of 3D images using primarily 2D slices. It's incredibly simple to use, but combines interaction, user-assignable encoding (using color), and small multiples very nicely for exploring the structure of 3D data in a 2D interface.
Brien - Mar 09, 2006 07:57:15 pm
Something that catches my attention in M. McGuffin et al is the performance mention in section 4. It says 500k voxels rendered on a 1.7GHz processor at 13fps, using no culling techniques (?). I also picked up that each voxel is rendered with an individual call to the graphics system. This works out to ~262 cycles per voxel (with a dedicated processor). I don't have much background in OpenGL/GLUT or volume rendering, but is this exceptionally good performance? I have more background with Java, where just rendering with individual calls, in any order, 13x500k small rectangles into spatially close memory takes ~9seconds on a 1.8GHz processor. I haven't run the same test using OpenGL. I can see how the octree could be used to speed up rendering. For example, if you label the center point of each octant with its distance from the viewer (using a simple vector projection), you could sort out the rendering order as you traverse.
I feel this paper didn't much say as much on the technical side as it could have. It seems they might be pulling some neat tricks, but maybe I just have my numbers out of scale.
Noaa - Mar 11, 2006 05:03:36 pm
I'm with Nchentan on the morbidity of the head-peeling in the paper. Since Prof. indicated that this is not even the way that a head is truly composed, I wonder 1) why they used that as their model, and 2) what could this actually be applied to? You know, what non-stomach-churning applications does this technology have...
Ashley - Mar 13, 2006 06:45:23 pm
The animated instructions reminded me a lot of popup books, or old hypercard games. Especially in hypercard games, being able to find hot-spots on the static images which allow for some sort of constrained movement or tiny animation is far more enjoyable than having a completely static image. It seems like these animated instructions would therefore be especially useful for kids. Not only is it more clear, but it's more interesting, and forces interaction with the user. I think because of this enjoyment/interaction factor, in addition to the clarity, I'd be more likely to remember/replicate what I'd seen.
IvanTam - Mar 14, 2006 11:36:41 am
I'm with Noaa on this one. The widgets that were presented in the "head" model seemed to be designed only for cutting apart heads (and doing so in the most morbid manner possible)? How useful are these tools for other applications? For example, how useful are these widgets for picking apart a model of a tractor? Is there are general set of widgets that these voxel modelling programs can converge on?
Raymond - Mar 14, 2006 01:31:43 pm
I thought the interactive exploded view was really cool and if you combine the automated assembly program that was presented in Feb. 28th's lecture of "Indentifying Design Principles" it would produce the best set of instructions. I know for me it helps a lot to see something being taken apart as well as seeing the cutways and sections to get the overall sense while trying to assemble my product.
Bryan - Mar 16, 2006 07:20:23 pm
Despite it's flaws, I did enjoy the novelty of the 3D exploration tools for the voxels. It seems like maybe a better application area for this work might be in exploring layers of the earths crust or even soil layers, which probably have a stronger ordering.
Gwyu - Mar 16, 2006 09:02:06 pm
The exploded view visualization reminds me a lot of the solid models that mechanical engineers use. I've seen non-interactive animations of the assembly and disassembly of models that are really compelling, but I'm not aware of any tool that takes the models and turns them into manipulable visualizations. I think some kind of player that would take a solid model and let users play with them would be interesting; to me at least, this seems feasible, especially in properly created models, the relationship/hierarchy among the parts is defined in the creation of the model.
Cynbot - Mar 18, 2006 09:37:33 pm
Gwyu Solidworks has this functionality and it is built into their workflow for designing assemblies. In the old days, Open Inventor had a cool viewer that would let you manipulate solid models.
One comment in regards to the lecture. The reason I didn't see the connection of the "head loaf" research to my work was that I was specifically building a surgical training environment. As we saw in class, most doctors perfered the more realistic drawings and my guess is that in those circumstances where much is at stake, you would like as many markers as possible. When I was learning to dissect cadavers and animal subjects, I wanted to make very sure I was following protocol. One wrong incision and you can cut through an artery or an intenstine or punture an organ.
Even though anatomy can differ from subject to subject, those differences aren't as drastic as one might think. So having these detailed images can relieve some anxiety in those situations. At least it did for me.
In the end, the system we created wound up being most useful for reconstructive surgery planning.
Mattkam - Mar 21, 2006 11:57:48 pm
The exploded view visualization could be extended for educational purposes if the various subparts of the model had certain relationships (or rules) that the designer wants the viewer to learn. In that case, this visualization could be used to encode such a simulation, i.e. think: microworld. But on top of simply allowing the viewer to play with these subparts, the viewer should perhaps be asked to achieve certain learning goals, e.g. assemble them into something.
