A New Approach to Affine Transform Invariant Shape Matching
In this paper a new approach to affine transform invariant matching of curvilinear contours is presented. In pattern recognition this matching problem has to be solved if plane objects arbitraily oriented in the 3D space are to be identified.
The approach is based on a new method of decomposing a shape into sections, which could be segmented invariant to affine transforms. For contour matching these sections of a curve will be used. These sections we will call form elements. For every contour an attributed relational model is constructed. Therefore the matching problem is reduced to that of comparing attributed relational descriptions. In a rough survey the matching procedure is shown. For that the best association of nodes is searched. This association allows to compute all parameters of the affine transform.
The feasibility of the approach is demonstrated for real plane objects.
KeywordsForm Element Match Problem Straight Line Segment Plane Object Contour Point
Unable to display preview. Download preview PDF.
- M.T. Miyatake, M. Nagao: Affine transform invariant curve recognition using Fourier descriptors, Trans. Inf. Process. Soc. Japan 24 no.1, 1983, pp. 64–71.Google Scholar
- J.K. Cheng, T.S. Huang: Recognition of Curvilinear Objects by Matching Relational Structures, Proc. Conf. on Pattern Recognition and Image Processing, 1982, pp. 343-348.Google Scholar
- Y. Lamdan, J.T. Schwartz and H.J. Wolfson: Object Recognition by Affine Invariant Matching, Proc. Conf. on Computer Vision and Pattern Recognition, 1988, pp. 335-344.Google Scholar
- S. Frydrychowicz: The Local Form — a New Method for Contour Segmentation, in V. Cantoni, L.P. Cordella, S. Levialdi and G. Sanniti di Baja (ed), Progress in Image Analysis and Processing, 1990, pp. 322-326.Google Scholar