I hold a Ph.D. in Computer Science from the University of Poitiers, France. I worked for the departement XLIM-SIC. Starting in January 2009, I was invited at the LIRIS, Lyon, in order to finish my research. I successfully defended my thesis on November 25th 2009 at the University of Poitiers.

Research interests

I am interested in the research areas of image analysis and topological modeling. During my Ph.D. thesis, I worked on 3D image processing using a topological model to propose new algorithms. My research focused on the segmentation of 3D image. I studied operations on the topological model to modify the represented partition of the image into regions. I used a cellular model to represent regions in 3D images. I developed incremental algorithms to compute topological features. That information is used as a criteria in the segmentation process. I worked on a new 3D deformable model based on simple-point that work in multi-label images. I am currently finishing work around a split and merge segmentation process using several topological criteria to segment 3D medical image.

In the future I am willing to look further into medical imaging applications as it represents a logical follow-up to my research. I am currently looking for a postdoctoral position in this domain.

Past and current works

During my master thesis, I worked on a topological model for image analysis : the topological map. My work involved the modification of an algorithm to extract a topological map from a 3D image since the existing algorithm only process labelled image. I also worked on a first approach of a bottom-up segmentation process using the topological map to represent images. I defended my master thesis at the University of Poitiers in July 2006.

I pursued these works during the first year of my PhD. I developped two approaches to merge regions in the topological representation of an image. This work have been presented at the 12th International Workshop on Combinatorial Image Analysis in Buffalo, NY, Appril 2008 and appears in the proceedings. I used these operation to provide a first approach to image segmentation using a classical color-based criterion. This work have been presented at the 14th Internationnal Conference on Discrete Geometry for Computer Imagery in Lyon, France, April 2008 and appears in the proceedings.

During the second year of my PhD, I worked toward using topological features of the regions captured by the topological map in order to provide new segmentation criteria. Thus, I developped algorithms to compute the Betti numbers, a well-known topological invariant, of regions in the topological map. This work have been presented at the First Workshop on Computational Topology in Image Context in Poitiers, France, June 2008. I used this work to incrementally compute Betti numbers during a region merging process to propose an operation that control the topology of the resulting partition. This work is beeing published in an issue of Discrete Applied Mathematics. I also started to work on the split of regions. I developped an algorithm that allow to split a region using a split complex that will give the shape of the resulting regions.

As a side project I work with Olivier Alata and Samuel Burg to provide them a functionnal prototype of the topological map allowing us to develop a novel segmentation algorithm based on Markov Random Field.

In my third year, I pursued works on the split region operation. I also worked on a new deformable model based on simple point that allow to deform surfaces in a multi-labelled images. This last feature is the key point of the new model as it allow to deform regions in the topological map to better fit the partition with image data and controling the topology of the partition. This work is currently under submission. I am currently working on the writting of my thesis and intend to finish work on the splitting algorithm over the summer.

Software

During my PhD, I developped a research software called topomap3d which now has a website (currently only in french). The software is not currently available for downloading but the source are available upon request to guillaume.damiand@liris.cnrs.fr.