Learning of Form Models from Exemplars
Model-based image recognition requires a general model of the object that should be detected. In many applications such models are not known a priori, but have to be learnt from examples. In this paper we describe our procedure for the acquisition and learning of general contour models. We developed a modified Procrustes algorithm for alignment and similarity calculation of shapes. Based on the calculated pair-wise similarity we learn groups of shapes. For each group we calculated prototypes. The set of prototypes will be used as models for the detection of object instances in new images.
KeywordsAlternaria Alternata Point Correspondence Contour Point Object Instance Boundary Pixel
- 1.Hill, A., Taylor, C.J., Brett, A.D.: A Framework for Automatic Landmark Identification Using a New Method of Nonrigid Correspondence. IEEE PAMI 22(3), 241–251Google Scholar
- 2.Zamperoni, P.: Methoden der digitalen Bildverarbeitung. Vieweg Verlag (1995)Google Scholar
- 3.Wall, K., Daniellson, P.-E.: A fast sequential method for polygonal approximation of digitized curves. Comput. Graph. Image Process 28, 220–227Google Scholar
- 5.Lele, S.R., Richtsmeier, J.T.: An Invariant Approach to Statistical Analysis of Shapes. Chapman & Hall/ CRC (2001)Google Scholar
- 7.Rangarajan, A., Chui, H., Bookstein, F.L.: The Softassign Procrustes Matching Algorithm. In: Proc. Information Processing in Medical Imaging, pp. 29–42 (1997)Google Scholar