contact    site map  
    polski   english


Marcin Janaszewski was awarded the title of Doctor of Computer Science in 2003. He is an employee of Computer Engineering Department since October 2005.
His scientific interest covers:

His training at ESIEE consisted in studying and applying in practice computer algorithms for 3D image analysis based on discreet topology. The main task consisted in building an original algorithm of hole filling in 3D volumetric objects. In the frame of the training he participated in three conferences and more than ten seminars. These seminars were organized by ESIEE.

As a result of the research Marcin Janaszewski in cooperation with prof. Michel Couprie (ESIEE) and dr Laurent Babout (KIS) created, implemented and tested the original algorithm of hole filling for volumetric objects were volumetric object is represented with a subset of voxels from a 3D image. The remaining voxels which does not belong to the object represent the background. These kind of images are usually results of tomography image segmentation.

An exemplary result of the algorithm when applied for simple volumetric object is presented in Fig 1.  The figure shows that the hole filling algorithm generates a volumetric object which not only close the corresponding hole but also locally interpolates the thickness of the input object.

Fig. 1. An example of hole filling volume for a torus. (a) Isosurface of a torus. (b) Torus with the hole filled. Hole filling volume is indicated with red color. (c) Isosurface of the torus hole filling volume. (d) Torus with the hole closed. Hole closing object indicated with red color is one voxel thick independently on the thickness of the input object.


The hole filling algorithm has the following properties:

In material science a hole filling algorithm can contribute to the quantification of damage phenomena that can help to understand and further optimize the resistance of the material to damage. Indeed, computed tomography allows reconstruction of 3D volumetric images which reveal microstructural details of a material [1]. Experiments on material damage processes can result in 3D images of crack propagation inside a material and interaction with the microstructure. For instance, in the case of stress corrosion cracking in austenitic stainless steel [2], a crack usually propagates through the weakest areas of the material sample, so if it meets the areas of stronger resistance, it tends to bypass the region by branching, subsequently rejoining and creating a hole of the crack path (see Fig. 2). Therefore, each hole in a crack corresponds to a material portion (also called ligament) of stronger resistance. The tendency for crack bridging is a measure of crack propagation resistance, and there is a need to quantify the distribution of ligaments. Segmentation of such portions is being done manually by visual inspection of 3D images. Unfortunately, such a strategy is time consuming and potentially unreliable [3], so an automatic system of hole segmentation for volumetric images seems to be a valuable alternative. Other possible medical applications may consist in filling small noisy holes in 3D tomography of human organs or segmentation of important holes in order to plan a surgery.

Fig. 2. Visualisation of a stainless steel sample and a crack. (a) Izosurface of stainless steel sample with a portion of crack on its outer surface. (b) Visualisation of crack propagation inside the sample;
 (c) Enlarged fragment of the crack from figure (b) where two bridges B1 and B2 are marked. (d) Enlarged view of bridge B1.


The algorithm has been tested on artificialy generated objects and very complicated 3D images which represent stress-corrosion crack in stainless steel. Fig 3 shows exemplary result for 3D image of a crack (stainless steel sample is represented by the background). Fig 3b shows that the algorithm filled each hole with an object (red color) which thickness locally match thickness of the crack neer the hole 


Fig 3. An example of hole filling result.  a) Crack visualisation, there are several holes of different size which represent bridges. c) The result of hole filling (red color) superimposed on the crack.

The result from fig 3 can be precisely observed in the following movie:

The outcome of the research has been published on conferences and in journals: [4-7].



[1] S.R. Stock, Recent advances in X-ray microtomography applied to materials, Int. Mater. Rev. 53 (3) (2008), pp. 129–181

[2] L. Babout, T. Marrow, D. Engelberg and P. Withers, X-ray microtomographic observation of intergranular stress corrosion cracking in sensitised austenitic stainless steel, Mater. Sci. Technol. 22 (2006), pp. 1068–1075.

[3] A. King, G. Johnson, D. Engelberg, W. Ludwig and J. Marrow, Observations of intergranular stress corrosion cracking in a grain-mapped polycrystal, Science 321 (2008), pp. 382–385.

[4] M. Janaszewski, M. Couprie, L. Babout, Hole filling in 3D volumetric objects, Pattern Recognition, vol. 43, pp. 3548-3559, 2010.

[5] M. Janaszewski, M. Couprie, L. Babout: Geometric Approach to Hole Segmentation and Hole Closing in 3D Volumetric Objects, Lecture Notes in Computer Science, vol. 5856, pp. 255-262, 2009.

[5] M. Janaszewski, L. Babout, M. Postolski, Zastosowanie algorytmów zamykania i wypełniania tuneli w komputerowej analizie materialów na bazie tomograficznych obrazów 3D, Scientific Notes of technical University of Łódź, Elektryka, vol. 121, pp. 159-172, 2010, (in polish).

[6] M. Janaszewski, L. Babout, M. Postolski, Ł Jopek, Segmentacja otworów w obiektach wolumetrycznych, Scientyfic Notes of AGH University of Science and Technology, pp. 855-864, 2009.

[7] M. Janaszewski, L. Babout, M. Postolski, Ł. Jopek, Zamykanie otworów w trójwymiarowych obiektach wolumetrycznych, Scientyfic Notes of AGH University of Science and Technology , Automatyka, pp. 865-878, 2009, (in polish).

Room: 318 (KIS - ul. Stefanowskiego 18/22)

Telephone: (+48)42 631 2750 (ext. 318)


return to the top

Laurent Babout

Last modification:
2010-11-16 14:42:04, Marcin Janaszewski