What do you mean by voxcel driven? Do you mean regular 3D representation or something else? Please clarify and may be able to help.

Thank you for the reply. I mean rendering in the volume (voxel) of objects.

I do not get the "driven" thing either. Does this means you want some 3D reconstruction in some discretized volume? Have look at voxel coloring technique.

From what you described it seems that you are more interested in the visualisation part of 3D reconstruction. To generate the 3D data from 2D images via voxel based approach, you may have a look at this paper for some information:

C. Leung, B. Appleton, M. Buckley, and C. Sun, Embedded Voxel Colouring with Adaptive Threshold Selection using Globally Minimal Surfaces, International Journal of Computer Vision, 99(2):215-231, September 2012. 

http://dx.doi.org/10.1007/s11263-012-0525-8

Other related approach can be found here:

http://vision.middlebury.edu/mview/

If you have a collection of 2D images (such as a TIFFstack) or other volumetric file formats such as .mrc, you can view them with most molecular viewers or with a variety of specialized and general 3D data viewers.  For example, UCSF Chimera is a molecular viewer particularly well adapted to viewing volumetric data sets: http://www.cgl.ucsf.edu/chimera/data/tutorials/maps08/volume-basics.html

Whereas programs like Osirix are optimized for viewing 3D medical data such as MRI and CT scans. 

NIH puts out a very powerful open-source generalized viewer that handles 3D volumes as well as many other types of data called imageJ: http://imagej.nih.gov/ij/features.html

You can use imageJ to convert 2D stacks to many different 3D formats, etc. and to run powerful surfacing, skeletonization and other types of analysis tools.

https://itunes.apple.com/app/id419227089?mt=8

Your question makes me think about "shape from silhouette" where you have a collection of images of a shape under different points of view, and you want to recover the 3D shape.

It can be done approximatively with this pipeline:

- recover the different positions of the camera (and some calibration data)

- segmentation of the images to locate the shape in each of them -> silhouettes

- shape from silhouettes

https://www.cvg.ethz.ch/teaching/compvis/2011/lecture/vision08.pdf

I don't know if there exists softwares with this pipeline (segmentation is already an hard task).

There are approaches in computer vision that use a quasi-volumetric representation to track 3D objects from 2D images captured by multiple cameras. Basically the objects of interest are approximated by a series of volumetric 'blobs', which are then fitted to the 2D images via optimization.

A group at my institute has some publications on this, here's the webpage if this sounds relevant to you:

http://gvv.mpi-inf.mpg.de/

Thank you the feedback from all of you.

What I was doing is to use the voxlap engine (http://advsys.net/ken/voxlap.htm) for rendering the voxel file directly.