Abstract:
Classification of plane contours (closed continuous curves) of object images is an important problem in the theory of image recognition. To describe a contour, a small number of its most informative points is chosen. As a rule, these points have a large value of a curvature estimate. Information on positional relationship of these points is transformed into vector representations; one of its important characteristics can be the center of masses. In this paper, we estimate a probability of rejection of the center of masses of the vector representation from it exact position in one sufficiently general probabilistic model of noiseness of an image contour.
Fulltext:
PDF file (220 kB)
