Ma photo

Nassim Jibai


Nassim Jibai is at the end of his thrid year of his PhD in Computer Science at UJF and working at INRIA Rhone-Alpes at Grenoble, France. He is supervised by Prof. Nicolas Holzschuch and co-advised by Dr. Cyril Soler . He is financed by Region Rhone-Alpes through the LIMA project.
On a personal level, Nassim is a sport fanatic and all sorts of out-doors activities. He is involve in all sorts of entertainment activities; theater, dancing, music, etc...

Nassim's current research interests are filtering 3D tomographic data, distributed system, and parallel computation.


Nassim is finishing his Thesis on filtering 3D tomography data.
Currently, he is writing his thesis. He is also looking and candidating for positions in either the research field or industry.
He is interested in solving problems that involves distributed systems and parallelism.

Kindly, find my C.V here.

Research themes

Using anisotropic diffusion we smooth 3D tomographic data at a user-defined scale while preserving the sharp features of the model.
All the computation is done on the GPU using CUDA as the subject fits the characteristics of embarrassingly parallel problems.
Here is an abstract of the research:

Three-dimensional volumes produced using tomographic reconstruction are inherently contaminated by noise that is introduced by the reconstruction algorithms.
When iso-contour surfaces are extracted from these volumes, the noise manifests as intricate geometric and topological artifacts in the surfaces.
We present a method to smooth the 3D tomographic data, before surface extraction, while preserving features at a specified scale.
Our algorithm is controlled using a single user parameter — the minimum scale of features to be preserved. Any variation that is smaller than the specified scale
is treated as noise and smoothed, while discontinuities such as corners, edges and detail at a larger scale are preserved. We demonstrate that our smoothed data
produces clean contour surfaces using standard surface-extraction algorithms. Our method is inspired by anisotropic diffusion within the volume. We compute our
diffusion tensors from the local continuous histograms of gradients around each voxel, thus extending previous approaches that use ad-hoc diffusion tensors.
Since our smoothing method works entirely on the GPU, it is extremely fast.

inhilator CT inhilator CT inhilator CT
                      (a)                                             (b)                                             (c)
inhilator CT inhilator CT inhilator CT
                      (d)                                             (e)                                             (f)
Slices of (a) the original CT reconstruction of the inhaler model, (b) close-up view of an area of (a)
squared in red and (c) the extracted surface of (a). (d) CT slice of the inhaler model after smoothing
by our method, (e) close-up view of the smooth CT (d) square in red, and (f) extracted surface from
smoothed data. We raised the gamma level in the CT images to better visualize the details of the volumes.

inhilator CT
Continuous local gradient histogram with their respective diffusion tensors at distinct locations. The top figures show
a histogram computed over a continuous region resulting with one dominant direction. Its respective diffusion tensor
is flat and orthogonal to the dominant direction. The bottom figures show a histogram computed over a region
containing an edge resulting with two dominant directions. Its respective diffusion tensor results in a cylinder shape
orthogonal to the dominant directions and parallel to the edge.

Alum300 boxes Alum300 k20 d40 boxes
                      (a )                                                                                         (b)                                            
Mechanical part (300^3 voxels): (a) surface extracted from original data and (b) surface extracted after smoothing by
our algorithm. Note how our algorithm removes the surface noise while preserving the sharp features such as the internal
threads or the letters on top of the model.

Recent publications



Lissage multi-echelle sur GPU des images et volumes avec preservation des details
Nassim Jibai
Mathématiques générales [math.GM]. Université de Grenoble, 2012. Français. ⟨NNT : 2012GRENM025⟩
Accès au texte intégral et bibtex BibTex


Communication dans un congrès

Multiscale Feature-Preserving Smoothing of Tomographic Data
Nassim Jibai, Cyril Soler, Kartic Subr, Nicolas Holzschuch
ACM SIGGRAPH 2011 Posters, ACM, Aug 2011, Vancouver, Canada. pp.Article No. 63, ⟨10.1145/2037715.2037786⟩
Communication dans un congrès
DOI : 10.1145/2037715.2037786
Audience internationale
Accès au texte intégral et bibtex BibTex

» See the complete list of my publications
» Image gallery

Curriculum Vitae

Resume (short version pdf)

Short Bio

2009-2012 PhD in Computer Graphics, Grenoble Universities, INRIA Rhône-Alpes - Grenoble, France.

2003-2006 Master's Degree in Computer Science, A library for implementing and using sequential and distributed recursive subdivision surfaces, AUB, Beirut, Lebanon. Find my Thesis Here in PDF.
This work was published in the Proceeding of Computer-Aided Design & Application 2006 entitled A Topological Abstraction for Implementing Recursive Surface Computations .

2000-2002 B.Sc in Computer Science; American University of Beirut, Beirut, Lebanon.


Instructor for Introduction to Programming: fall 2003-2004 and fall 2007-2008.

At AUB; American University of Beirut, Lebanon

Personal area

Here is what I like. Some links.

Some links...
» Related pages
» Cheap solar system travel