Classification of 2D grayscale objects in a space of the multiresolution representations
Sergey Ganebnykh, M. M. Lange · Pattern Recognition and Image Analysis · 2009
A new approach is developed for fast voting-based classification of 2D patterns given by the grayscale images and represented by the trees of elliptic primitives. Due to a multiresolution property of the representations, the fast search algorithm is suggested to make voting-based decisions about the classes of the submitted objects and it is shown that this algorithm requires much smaller computations as compared with a full search algorithm for the decision. A computational complexity of the fast algorithm is O ( K log K ) when a number of the classes K is large. An efficiency of the proposed approach is shown by experimental results on signature and hand gesture recognition in terms of the error rate as the function of the resolution level.