Hi everybody,
do you know a good entry-point/overview over super resolution algorithms for 3D data (also implementations/libraries are highly welcome)?
I want to produce highly accurate (groundtruth-like) 3D models from multiple partial pointclouds.
Any help appreciated....
Regards,
Bernd
Hi,
You didn't say what the models would be (field) so I'm guessing you just want a general purpose start.
Ground truth quality might be a bit of an exaggeration. Not even Hollywood productions produce that kind of quality much less in common scientific 3D models.
The accuracy of your model is only guaranteed in your control points, the rest is necessarily interpolated or estimated. So generally speaking your model will only be as accurate as the number of points you've sampled (whatever happens in between them is speculated no matter how good a speculation it is).
That said there have been good experiments made in high rendering engines. My suggestion is that you start with blender: http://www.blender.org/
It's free, open-source and with it's own Python API. So it's great for tweaking and customizing. Here's one project that attempts something similar (or at least I think it's similar) to your request: http://www.bioblender.eu/
Be prepared to deal with a very time consuming and machine intensive process for rendering images and animations. If you're having problems dealing with the data size you can consider Visit (works in parallel or MPI): https://wci.llnl.gov/simulation/computer-codes/visit/
Visit deals with a lot of data but I think it's rendering capabilities (textures, light, etc.) are far less than blender. It all depends on your final objective.
Hope it helps.
Hello Bernd,
Please go through herewith the attached paper may helpful to you
This is what you are looking for. These researchers worked for human bodies, but you might use it for other bodies, I guess.
Y. Cui, W. Chang, T. Nöll, y D. Stricker, «KinectAvatar: Fully Automatic Body Capture Using a Single Kinect», Computer Vision - ACCV 2012 Workshops, vol. 7729, J.-I. Park y J. Kim, Eds. Springer Berlin Heidelberg, 2013.
Thank you all for the helpful hints, offering a good starting point!
I'll check them in detail and build up on them.
@Rhadames: 3 of the 4 authors of your proposed work are members of my research group. I'm therefore aware of this publication, but I'll let them know, that other researchers consider their work to be relevant as well... ;)
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Computer Graphics and Animation