Abstract
The main contribution of the present article consists of new 3D parallel and symmetric thinning schemes which have the following qualities:
-
They are effective and sound, in the sense that they are guaranteed to preserve topology. This guarantee is obtained thanks to a theorem on critical kernels;
-
They are powerful, in the sense that they remove more points, in one iteration, than any other symmetric parallel thinning scheme;
-
They are versatile, as conditions for the preservation of geometrical features (e.g., curve extremities or surface borders) are independent of those accounting for topology preservation;
-
They are efficient: we provide in this article a small set of masks, acting in the grid ℤ3, that is sufficient, in addition to the classical simple point test, to straightforwardly implement them.
Similar content being viewed by others
References
Aktouf, Z., Bertrand, G., Perroton, L.: A three-dimensional holes closing algorithm. Pattern Recognit. Lett. 23(5), 523–531 (2002)
Alexandroff, P.: Diskrete Räume. Mat. Sb. 2, 501–518 (1937)
Alexandroff, P., Hopf, H.: Topologie, I. Springer, Berlin (1935)
Bernard, T., Manzanera, A.: Improved low complexity fully parallel thinning algorithm. In: Proceedings 10th International Conference on Image Analysis and Processing (ICIAP’99) (1999)
Bertrand, G.: Simple points, topological numbers and geodesic neighborhoods in cubic grids. Pattern Recognit. Lett. 15, 1003–1011 (1994)
Bertrand, G.: On P-simple points. C. R. Acad. Sci., Ser. 1 Math. 321, 1077–1084 (1995)
Bertrand, G.: New notions for discrete topology. In: Discrete Geometry for Computer Imagery. LNCS, vol. 1568, pp. 218–228. Springer, Berlin (1999)
Bertrand, G.: On critical kernels. C. R. Acad. Sci., Ser. 1 Math. I(345), 363–367 (2007)
Bertrand, G., Aktouf, Z.: A three-dimensional thinning algorithm using subfields. In: Vision Geometry III, vol. 2356, pp. 113–124. SPIE, Bellingham (1996)
Bertrand, G., Couprie, M.: A new 3D parallel thinning scheme based on critical kernels. In: Discrete Geometry for Computer Imagery. LNCS, vol. 4245, pp. 580–591. Springer, Berlin (2006)
Bertrand, G., Couprie, M.: Two-dimensional parallel thinning algorithms based on critical kernels. J. Math. Imaging Vis. 31(1), 35–56 (2008)
Bertrand, G., Couprie, M.: On parallel thinning algorithms: minimal non-simple sets, P-simple points and critical kernels. J. Math. Imaging Vis. 35(1), 23–35 (2009)
Bertrand, G., Malandain, G.: A new characterization of three-dimensional simple points. Pattern Recognit. Lett. 15(2), 169–175 (1994)
Bing, R.: Some aspects of the topology of 3-manifolds related to the Poincaré conjecture. In: Lectures on Modern Mathematics, vol. II, pp. 93–128 (1964)
Blum, H.: A Transformation for Extracting New Descriptors of Shape (1967)
Couprie, M.: Note on fifteen 2D parallel thinning algorithms. Internal report, Université de Marne-la-Vallée, IGM2006-01 (2005)
Couprie, M., Bertrand, G.: New characterizations of simple points in 2D, 3D and 4D discrete spaces. IEEE Trans. Pattern Anal. Mach. Intell. 31(4), 637–648 (2009)
Duda, O., Hart, P., Munson, J.: Graphical data processing research study and experimental investigation. Tech. Rep. AD650926 (1967)
Giblin, P.: Graphs, Surfaces and Homology. Chapman & Hall, London (1981)
Golay, J.: Hexagonal parallel pattern transformations. IEEE Trans. Comput. C-18, 733–740 (1969)
Kong, T.: On the problem of determining whether a parallel reduction operator for n-dimensional binary images always preserves topology. In: Procs. SPIE Vision Geometry II, vol. 2060, pp. 69–77 (1993)
Kong, T.Y.: On topology preservation in 2-D and 3-D thinning. Int. J. Pattern Recognit. Artif. Intell. 9, 813–844 (1995)
Kong, T.Y.: Topology-preserving deletion of 1’s from 2-, 3- and 4-dimensional binary images. In: Discrete Geometry for Computer Imagery. LNCS, vol. 1347, pp. 3–18. Springer, Berlin (1997)
Kong, T.Y.: Minimal non-simple and minimal non-cosimple sets in binary images on cell complexes. In: Discrete Geometry for Computer Imagery. LNCS, vol. 4245, pp. 169–188. Springer, Berlin (2006)
Kong, T.Y., Rosenfeld, A.: Digital topology: introduction and survey. Comput. Vis. Graph. Image Process. 48, 357–393 (1989)
Kovalevsky, V.: Finite topology as applied to image analysis. Comput. Vis. Graph. Image Process. 46, 141–161 (1989)
Lohou, C.: Contribution à l’analyse topologique des images: étude d’algorithmes de squelettisation pour images 2D et 3D, selon une approche topologie digitale ou topologie discrète. Ph.D. thesis, Université de Marne-la-Vallée, France (2001)
Lohou, C.: Detection of the non-topology preservation of ma’s 3D surface-thinning algorithm, by the use of P-simple points. Pattern Recognit. Lett. 29, 822–827 (2008)
Lohou, C., Bertrand, G.: Two symmetrical thinning algorithms for 3D binary images. Pattern Recognit. 40, 2301–2314 (2007)
Ma, C.: On topology preservation in 3D thinning. Comput. Vis. Graph. Image Process. 59(3), 328–339 (1994)
Ma, C.M.: A 3D fully parallel thinning algorithm for generating medial faces. Pattern Recognit. Lett. 16(1), 83–87 (1995)
Ma, C.M., Sonka, M.: A 3D fully parallel thinning algorithm and its applications. Comput. Vis. Image Underst. 64(3), 420–433 (1996)
Ma, C.M., Wan, S.Y., Lee, J.D.: Three-dimensional topology preserving reduction on the 4-subfields. IEEE Trans. Pattern Anal. Mach. Intell. 24(12), 1594–1605 (2002)
Malgouyres, R., Francés, A.: Deciding whether a simplicial 3-complex collapses to a 1-complex is NP-complete. In: Discrete Geometry for Computer Imagery. LNCS, vol. 4992, pp. 177–188. Springer, Berlin (2008)
Manzanera, A., Bernard, T., Prêteux, F., Longuet, B.: Medial faces from a concise 3D thinning algorithm. In: Proceedings IEEE International Conference on Computer Vision (ICCV’99), pp. 337–343 (1999)
Manzanera, A., Bernard, T., Prêteux, F., Longuet, B.: Ultra-fast skeleton based on an isotropic fully parallel algorithm. In: Proceedings Discrete Geometry for Computer Imagery (DGCI’99). Lecture Notes in Computer Science, vol. 1568, pp. 313–324. Springer, Berlin (1999)
Manzanera, A., Bernard, T., Prêteux, F., Longuet, B.: n-dimensional skeletonization: a unified mathematical framework. J. Electron. Imaging 11(1), 25–37 (2002)
Németh, G., Kardos, P., Palágyi, K.: Topology preserving 2-subfield 3D thinning algorithms. In: Signal Processing, Pattern Recognition and Applications (SPPRA 2010), vol. 678, pp. 311–316 (2010)
Németh, G., Kardos, P., Palágyi, K.: Topology preserving 3D thinning algorithms using four and eight subfields. In: Campilho, A., Kamel, M. (eds.) Image Analysis and Recognition. Lecture Notes in Computer Science, vol. 6111, pp. 316–325. Springer, Berlin (2010)
Palágyi, K.: A 3D fully parallel surface-thinning algorithm. Theor. Comput. Sci. 406(1–2), 119–135 (2008)
Pfaltz, J., Rosenfeld, A.: Computer representation or planar regions by their skeletons. Commun. ACM 10, 119–125 (1967)
Ronse, C.: Minimal test patterns for connectivity preservation in parallel thinning algorithms for binary digital images. Discrete Appl. Math. 21(1), 67–79 (1988)
Rosenfeld, A.: Connectivity in digital pictures. J. ACM 17, 146–160 (1970)
Rosenfeld, A.: A characterization of parallel thinning algorithms. Inf. Control 29(3), 286–291 (1975)
Rosenfeld, A., Pfaltz, J.: Sequential operations in digital picture processing. J. ACM 13, 471–494 (1966)
Saha, P., Chaudhuri, B., Chanda, B., Dutta Majumder, D.: Topology preservation in 3D digital space. Pattern Recognit. 27, 295–300 (1994)
Whitehead, J.: Simplicial spaces, nuclei and m-groups. Proc. Lond. Math. Soc. 45(2), 243–327 (1939)
Zeeman, E.: On the dunce hat. Topology 2, 341–358 (1964)
Author information
Authors and Affiliations
Corresponding author
Additional information
This work has been partially supported by the “ANR-2010-BLAN-0205 KIDICO” project.
Appendix
Appendix
Proof of Theorem 5
The following result is a consequence of Theorem 4.3(iii) of [8], note that this theorem holds for complexes of arbitrary dimensions.
Let \(S \in \mathbb{X}^{3}\) , let R⊆S, and let T such that R⊆T⊆S. If R − contains the critical kernel of S, then T − collapses onto R − .
Now let \(X \in \mathbb{X}^{3}\) and let Y⊆X such that Y − contains the critical kernel of X.
Let X∖Y={x 1,…,x k }. Thus the faces of X∖Y are ordered according to their indices in an arbitrary way.
We set X 0=X, X i =X∖{x 1,…,x i }, i∈[1,k].
Let i∈[1,k]. The complex X i contains Y, thus \(X_{i}^{-}\) contains all the faces which are critical for X. By the above result \(X_{i-1}^{-}\) collapses onto \(X_{i}^{-} = [X_{i-1} \setminus \{ x_{i}\}]^{-}\), which means that x i is simple for X i−1 and that X i is an elementary thinning of X i−1 (Definition 1). Thus, the xel complex Y=X k is a thinning of X=X 0. □
Proof of Theorem 7
Let \(X \in \mathbb{X}^{3}\) and let Y⊆X.
-
(i)
Suppose Y − contains the critical kernel of X. Let Z such that Y⊆Z⊆X. Since Z − contains the critical kernel of X, by Theorem 5, Z is a thinning of X.
-
(ii)
Suppose Y − does not contain the critical kernel of X. Then, there exists a face which is critical for X in X −∖Y −. There exists also a face x in X −∖Y − which is \(\mathcal{M}\)-critical for X. Then, the \(\mathcal{M}\)-crucial clique \(C = x^{+}_{X}\) is non-simple for X (see [12], Theorem 28, and Remark 19), i.e., the set Z=X∖C is not a thinning of X. We observe that Y⊆Z. Thus, there exists Z such that Y⊆Z⊆X, and such that Z is not a thinning of X.
□
Proof of Proposition 8
We proved the proposition with the help of a computer program. All 226 possible configurations of the neighborhood of a point x in X were examined, and for each of them the equivalence between Definition 1 and conditions (1) and (2) was successfully tested. □
Proof of Proposition 10
We proved the proposition with the help of a computer program. All 216 possible configurations of the \(\mathcal{K}\)-neighborhood of a 2-clique C in X were examined, and for each of them the equivalence between Definition 3 and conditions (1) and (2) was successfully tested. □
Proof of Proposition 11
We proved the proposition with the help of a computer program. All 28 possible configurations of the \(\mathcal{K}\)-neighborhood of a 1-clique C in X were examined, and for each of them the equivalence between Definition 3 and the condition was successfully tested. □
Proof of Theorem 16 and Theorem 15
We proved the proposition with the help of a computer program. The condition “\(\mathcal{K}^{*}(C) \cap X\) is reducible” could not be checked directly because of combinatorial explosion, so we proved the property recursively with respect to the cardinality of \(S = \mathcal{K}^{*}(C) \cap X\). More precisely, knowing that the proposition is trivially true for |S|=0, we checked it for all possible configurations of n elements of S, for n=1,…,N (with N=26,16,8 for d=3,2,1 respectively, C being a d-clique), based on the fact that the proposition was already proved for n−1. For simplicity, the configurations of n elements out of N were generated by scanning all possible 2N configurations and selecting those with precisely n elements. □
Proof of Theorem 17
We proved the theorem with the help of a computer program. It is trivially true when |S|=0 (case of a 0-clique). We checked it for all possible configurations of N elements of S, (with N=26,16,8 for d=3,2,1 respectively, C being a d-clique). For each of these configurations, we tested for each simple voxel x of S the reducibility of S∖{x}, thanks to Theorem 16 and to characterizations of regular cliques (Proposition 13) and simple points (Proposition 8). □
Proof of Theorem 21
Let \(X \in \mathbb{V}^{3}\), let C be a d-clique which is critical for X, and let x=⋂{x∈C}.
Suppose C is not \(\mathcal{M}\)-crucial for X. Then there exists a d′-clique D which is critical for X, and such that x is a proper face of the face y=⋂{x∈D}. Thus, we have d<d′ and D is a proper subset of C.
-
(i)
Suppose D is \(\mathcal{D}\)-crucial for X. It means that C contains a voxel belonging to a d′-clique which is \(\mathcal{D}\)-crucial for X, with d′>d. Thus, C is not \(\mathcal{D}\)-crucial for X.
-
(ii)
Suppose D is not \(\mathcal{D}\)-crucial for X. It means that D (hence also C) contains a voxel belonging to a d″-clique which is \(\mathcal{D}\)-crucial for X, with d″>d′>d. Again, C cannot be \(\mathcal{D}\)-crucial for X.
Thus, a clique which is \(\mathcal{D}\)-crucial for X is necessarily \(\mathcal{M}\)-crucial for X. It follows that the \(\mathcal{D}\)-crucial kernel of X is a subset of its \(\mathcal{M}\)-crucial kernel. □
Rights and permissions
About this article
Cite this article
Bertrand, G., Couprie, M. Powerful Parallel and Symmetric 3D Thinning Schemes Based on Critical Kernels. J Math Imaging Vis 48, 134–148 (2014). https://doi.org/10.1007/s10851-012-0402-7
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10851-012-0402-7