On the Behavior of Spatial Critical Points under Gaussian Blurring A Folklore Theorem and Scale-Space Constraints
The main theorem we present is a version of a “Folklore Theorem” from scale-space theory for nonnegative compactly supported functions from ℝn to ℝ. The theorem states that, if we take the scale in scale-space sufficiently large, the Gaussian-blurred function has only one spatial critical extremum, a maximum, and no other critical points.
a sharp estimate, in terms of the radius of the support, of the scale after which the set of critical points consists of a single maximum;
all critical points reside in the convex closure of the support of the function
These results show, for example, that all catastrophes take place within a certain compact domain determined by the support of the initial function and the estimate mentioned in 1.
To illustrate that the restriction of nonnegativity and compact support cannot be dropped, we give some examples of functions that fail to satisfy the theorem, when at least one assumption is dropped.
Keywords and PhrasesLarge-scale behavior loss of detail nonnegative function compact support spatial critical point deep structure.
Unable to display preview. Download preview PDF.
- 2.J.A. Dieudonné. Foundations of Modern Analysis, volume 10 of Pure and Applied Mathematics. Academic Press, New York. London, 1960.Google Scholar
- 3.L.M.J. Florack. Image Structure, volume 10 of Computational Imaging and Vision. Kluwer Academic Publishers, Dordrecht. Boston. London, 1997.Google Scholar
- 5.B. Grünbaum. Convex Polytopes, volume 16 of Pure and Applied Mathematics. Interscience Publishers, London, 1967.Google Scholar
- 6.P. Johansen, S. Skelboe, K. Grue, and J. Damgaard Andersen. Representing signals by their toppoints in scale-space. In Proceedings of the 8th International Conference on Pattern Recognition, pages 215–217, Paris, 1986.Google Scholar
- 8.T. Lindeberg and B.M. Ter Haar Romeny. Linear Scale-Space I: Basic Theory, volume 1 of Computational Imaging and Vision, chapter 1, pages 1–38. Kluwer Academic Publishers, Dordrecht. Boston. London, 1994.Google Scholar