Symmetries of differential equations in computer vision tasks
Vrazhnov D.A., Shapovalov A.V., Nikolaev V.V.
Компьютерные исследования и моделирование Ижевск: Научно-издательский центр "Регулярная и хаотическая динамика", 2010 (in press). Vol. 2 (4). (accepted)
In our work we present generalization of well-known approach for constructing invariant feature vectors of images in computer vision tasks. Basic step of suggested algorithm is replacement of commonly used Gaussian filter by convolution of function of image with Green’s function of evolution operator, which inherits symmetries of the one operator. The use of general filtration allows computing additional characteristics of invariant feature vectors.
|
