I loaded a MRC file, isosurfaced it, gave it a transparency around 0.48, and generated a vrml in chimera. I loaded the vrml into amira. initial comparison: I reloaded the vrml into chimera, I turned off the fog on chimera, moving the object around the refresh rate seems to be around 8 fps. On amira it seems to be around 24fps. results: The graphics on Chimera seems to be better during the rotation of transparent objects: full rotation in chimera is flawless, full rotation in amira results in flipping of the object (vertices in back quantum-ly jump to the front, front vertices go to back, looks like a z buffer error) periodically at 180 degrees, making it useless for many camera positions. Any comments or suggestions?
On Wed, 2 Jul 2008, Matthew Dougherty wrote:
I loaded a MRC file, isosurfaced it, gave it a transparency around 0.48, and generated a vrml in chimera. I loaded the vrml into amira.
initial comparison: I reloaded the vrml into chimera, I turned off the fog on chimera, moving the object around the refresh rate seems to be around 8 fps. On amira it seems to be around 24fps.
results: The graphics on Chimera seems to be better during the rotation of transparent objects: full rotation in chimera is flawless, full rotation in amira results in flipping of the object (vertices in back quantum-ly jump to the front, front vertices go to back, looks like a z buffer error) periodically at 180 degrees, making it useless for many camera positions.
Any comments or suggestions?
What's wrong with just using chimera? If it's just that it's too slow, then please file a bug report and provide a sample data file. Typical transparency implementations require that the graphics primitives be drawn in a back-to-front order. It sounds like amira is just drawning the primitives in the order in which they are given in the vrml file, so the transparency effect is haphazard. You would probably get better results from a dedicated vrml viewer, like Octaga, BS Contact, Flux, FreeWrl, etc. Chimera uses an alternate, 2-pass, transparency technique that works well, but is not perfect (except within surfaces, where chimera does depth-sort the triangles). Greg Couch UCSF Computer Graphics Lab
Hi Matt, First, Chimera VRML rendering of transparent objects is wrong and generally not useful. It simply adds the brightness of all overlying surface layers. Correct behavior dims the appearance of layers in back because they are viewed through the layers in front. The wrong behavior gives no cues as to what is in front or in back and results in saturated full intensity colors when surfaces are more opaque than transparent. The Chimera implementation is just a cheap approximation to correct transparency. The code would have to be much more complex to draw the transparency correctly. Given that Chimera ignores the depth of VRML surface layers I'm surprised it renders slowly. How does the rendering speed of the original volume data in Chimera compare? I would expect that to be slower because it actually resorts the surface triangles by depth each time they are drawn. You can check exact frame rates using the Chimera benchmark tool, menu entry Tools / Utilities / Benchmark. I have plans to optimize the rendering of transparent volume surfaces -- I think it could be nearly as fast as opaque surfaces with some fancier code. It is unlikely we will make our VRML drawing handle transparency correctly since VRML viewing is not a primary function of Chimera. It might be possible to change the VRML implementation to use our volume / molecule surface code but many other Chimera projects have higher priority. Tom
What's wrong with just using chimera? If it's just that it's too slow, then please file a bug report and provide a sample data file.
First of all I think Chimera is a great package and I highly recommend it, regularly. It is one of the very few softwares that actually adheres to specifications correctly (eg VRML). Not complaining about the speed, it is more than adequate for animation. Appreciate the tradeoff on speed & quality, not clear on how you are doing it in opengl. no bugs to report.
after years of seeing most viz packages fail the same way on the opengl, I am impressed that that you got it the best I have seen, short of ray tracing. just surprised the commercial packages (at least 5 come to mind) that can't keep up.
as for using chimera exclusively, I have to integrate software (3dviz, photoshop, FCP NLE) depending on what users give me to start with. be nice just to use only one... I just try to figure out who is the best at what they are doing, and get data in and out of it.
thanks, Matt
Typical transparency implementations require that the graphics primitives be drawn in a back-to-front order. It sounds like amira is just drawning the primitives in the order in which they are given in the vrml file, so the transparency effect is haphazard. You would probably get better results from a dedicated vrml viewer, like Octaga, BS Contact, Flux, FreeWrl, etc. Chimera uses an alternate, 2-pass, transparency technique that works well, but is not perfect (except within surfaces, where chimera does depth-sort the triangles). Greg Couch UCSF Computer Graphics Lab
participants (4)
-
Dougherty, Matthew T. -
Greg Couch -
Matthew Dougherty -
Tom Goddard