ScaleSpaceViz: α-scale spaces in practice

F.M.W. Kanters, L.M.J. Florack, R. Duits, B. Platel, B.M. Haar Romenij, ter

Onderzoeksoutput: Bijdrage aan tijdschriftTijdschriftartikelAcademicpeer review

6 Citaten (Scopus)

Samenvatting

Kernels of the so-called a-scale space have the undesirable property of having no closed-form representation in the spatial domain, despite their simple closed-form expression in the Fourier domain. This obstructs spatial convolution or recursive implementation. For this reason an approximation of the 2D a-kernel in the spatial domain is presented using the well-known Gaussian kernel and the Poisson kernel. Experiments show good results, with maximum relative errors of less than 2.4%. The approximation has been successfully implemented in a program for visualizing a-scale spaces. Some examples of practical applications with scale space feature points using the proposed approximation are given.
Originele taal-2Engels
Pagina's (van-tot)106-116
TijdschriftPattern Recognition
Volume17
Nummer van het tijdschrift1
DOI's
StatusGepubliceerd - 2007

Vingerafdruk Duik in de onderzoeksthema's van 'ScaleSpaceViz: α-scale spaces in practice'. Samen vormen ze een unieke vingerafdruk.

  • Citeer dit