1. bookVolume 80 (2021): Edition 3 (December 2021)
Détails du magazine
License
Format
Magazine
eISSN
1338-9750
Première parution
12 Nov 2012
Périodicité
3 fois par an
Langues
Anglais
access type Accès libre

Improvement and Handling of the Segmentation Model with an Inflation Term

Publié en ligne: 01 Jan 2022
Volume & Edition: Volume 80 (2021) - Edition 3 (December 2021)
Pages: 119 - 134
Reçu: 08 Nov 2020
Détails du magazine
License
Format
Magazine
eISSN
1338-9750
Première parution
12 Nov 2012
Périodicité
3 fois par an
Langues
Anglais
Abstract

The use of balloon models to address the problems of “snakes” based models was introduced by Laurent D. Cohen. This paper presents a geodesic active contours model with a modified external force term that includes a balloon model. This balloon model makes the segmentation surface to behave like a balloon inflated by the external forces. In this paper, we show an automatic way to control the behaviour of the external force with respect to the segmentation evolution. The external forces, comprised of edge and inflation terms, push the segmentation surface to edges, while curvature regularizes the evolution. As segmentation evolves, the influence of the applied inflation force is determined by how close we are to the edges. With this setup, the initial segmentation does not need to be close to the object’s edges, instead it is inflated by the balloon model towards the edges. Closer to the edges, the influence of the inflation force is adjusted accordingly. The force’s influence is completely turned off when the evolution is stable (reached the edges), then only the curvature and edge information is used to evolve the segmentation.

This approach solves the issues associated with inclusion of balloon model. These issues are that the inflation force can overpower forces from weak edges, or they can cause the contour to be slightly larger than the actual minima. We present examples of the improved model for segmentation of human bladder images. Weak edges are more prevalent in medical images, and the automated handling of the inflation forces gives promising results for this kind of images.

Keywords

[1] ATKINSON, K.: An Introduction to Numerical Analysis, 2 edition, John Wiley & Sons, Inc. New York, 1989. Search in Google Scholar

[2] CASELLES, V.—KIMMEL, R.—SAPIRO, G.: Geodesic active contours, Internat. J. Comput. Vision 22 (1997), no. 1, 61–79. Search in Google Scholar

[3] COHEN, L. D.: On active contour models and balloons, CVGIP: Image Understanding 53 (1991), no. 2, 211–218. Search in Google Scholar

[4] DRBLÍKOVÁ, O.—MIKULA, K.: Semi-implicit diamond-cell finite volume scheme for 3D nonlinear tensor diffusion in coherence enhancing image filtering (Robert Ey-mard et al. eds.) In: Proceedings of the 5th International Symposium, Aussois, June, 2008, Finite Volumes for Complex Applications Vol. V: Problems and Perspectives ISTE and WILEY, London, 2008, pp. 343–350. Search in Google Scholar

[5] EYMARD, R.—HANDLOVIČOVÁ, A.—MIKULA, K.: Study of a finite volume scheme for the regularized mean curvature flow level set equation, IMA J. Numer. Anal. 31 (2011), no. 3, 813–846. Search in Google Scholar

[6] FROLKOVIČ, P.—MIKULA, K.: Flux-based level set method: A finite volume method for evolving interfaces, Appl. Numer. Math. 57 (2007), no. 4, 436–454. Search in Google Scholar

[7] KASS, M.—WITKIN, A.—TERZOPOULOS, D.: Snakes: Active contour models, Internat. J. Comput. Vision 1 (1988), 321–331.10.1007/BF00133570 Search in Google Scholar

[8] MIKULA, K.—REMEŠÍKOV Á, M.: Finite volume schemes for the generalized subjective surface equation in image segmentation, Kybernetika, 45 (2009), no. 4, 646—656. Search in Google Scholar

[9] MIKULA, K.—SMÍŠEK, M.—ŠPIR, V.: Parallel algorithms for segmentation of cellular structures in 2D+time and 3D morphogenesis data. In: Proceedings of the Conference Algoritmy, Vol. 2012, pp. 416–426. Search in Google Scholar

[10] URBÁN, J.: The new improvements of Atlas based image segmentation, Slovak University of Technology in Bratislava, Reg. no.: SvF-10911-38875 (2016), 47–50. Search in Google Scholar

Articles recommandés par Trend MD

Planifiez votre conférence à distance avec Sciendo