Issac Weiss,
Noise-resistant invariants of curves,
In Joseph Mundy and Andrew Zisserman eds., Geometric Invariance in Computer Vision,
Cambridge, MA, USA, Series: Artificial intelligence, pp. 135-156, The MIT Press, 1992.
Issac Weiss,
Noise-resistant invariants of curves,
In Joseph Mundy and Andrew Zisserman eds., Geometric Invariance in Computer Vision,
Cambridge, MA, USA, Series: Artificial intelligence, pp. 135-156, The MIT Press, 1992.In this chapter we concentrate on local or differential invariants of curves. Unlike global (algebraic) invariants, they do not suffer from the occlusion problem. A representation of a general curve is obtained which is independent of the viewpoint from which the curve is seen. This ``signature'' can be used to match an observed shape to one stored in a database, without having to search for the correct viewpoint. Both projective and affine invariants are treated. The reliability is improved in two ways: (1) We have found new methods for obtaining differential invariants which are more robust. (2) We have developed differentiation techniques which give much more reliable results than previous ones. These differentiation methods are useful in many other applications as well.