J-L. Rose, C. Muller, C. Odet
Purpose and Context
Image segmentation through region growing is a method that localizes and extracts an object from an image. When difficult segmentation context are linked to noise or a lack of information, the use of prior knowledge in the region growing process improves segmentation accuracy. Imaging of living is often affected by this context. Moreover, biomedical applications deal with large amounts of data. Then, it is mandatory to use a robust and fast processing. This context leads us to propose a shape prior for the region growing process. In [ROSE-08], we present our work composed of two main contributions: the Variational Region Growing and the integration of shape prior in Region growing.
Methods
Variational region growing
We proposed to define region growing in the variational framework. We chose to describe our region by a discrete function. The region evolution equation is defined in a variational formulation. An energy functional minimization is associated with our function. For the criterion minimum, the segmented region converges to the desired object. The main interest of our region growing approach is the use of a region-based energy minimization in the evolution process [ROSE-09].
Shape prior in region growing
We defined a region-based shape constraint in the segmentation. We introduced shape descriptors based on Tchebychev moments. We proposed to introduce a weighting term into our functional. This term allows to consider the Tchebychev moments hierarchy, thus adapting our criterion with the encoded moments information. We proposed a shape prior which incorporates invariance with respect to the group of similarity transformations of a given shape. Our shape constraint was defined for two-dimensional and three-dimensional images .
Bunny image | Region growing segmentation | Region growing segmentation |
We also proposed another region-based energy to integrate shape prior in region growing. This work relies upon a functional computed from the normalised signed distance between the evolving region and a registered reference model.
This criterion was applied in the framework of small animal imaging. 3D in-vivo μ-CT images of mouse kidneys and mouse skulls were segmented previously to a phenotyping analysis of biologists.
Initial μ-CT images of mouse kidneys
Reference volume
Region growing without shape prior
=> Leakages outside the kidney
Region growing with shape prior
=> Correct segmentation
Illustration of the variational region growing using shape prior constraint
Conclusion
We applied shape prior criteria to the variational region growing. We showed the ability of our method to incorporate region-based energy. We studied the ability of our shape constraints to drive the evolution of the region growing toward the desired shape. Segmentation results were given for synthetic and biomedical images.
Bibliography
[ROSE-09] J.L. Rose, C Revol-Muller, and C. Odet. New region growing based on a variational approach. In International Conference on Computer Vision Theory and Applications, Lisboa, Portugal, pages in-press, February 2009.
[ROSE-08b] J.L. Rose, C. Revol-Muller, J.B. Langlois, M. Janier, and C. Odet. 3D region growing integrating adaptive shape prior. In Biomedical Imaging: From Nano to Macro, 2008. ISBI 2008. 5th IEEE International Symposium on, Paris, France, pages 967-970, May 2008.
[ROSE-07] J.L. Rose, C. Revol-Muller, M. Almajdub, E. Chereul, and C. Odet. Shape Prior in an Automated 3D Region Growing method. In IEEE International Conference on Image Processing ICIP'07, San Antonio, USA, pages 53-56, September 2007.
Thesis
[ROSE-08a] J.L. Rose. Croissance de région variationnelle et contraintes géométriques tridimensionnelles pour la segmentation d'image. phdDoctorat, INSA Lyon, 2008. Jury: M. Revenu (rap.), F.Truchetet (rap.),P.Bolon (prés.), M.Jourlin , C.Revol-Muller (dir.),C.Odet(dir.).