Abstract: The work presents a method for extraction of a thin regularized contour enabling simultaneous processing of the scene images made in various spectral ranges. The method presented can be easily accomplished in the form of a cellular neural network. The basis of the method is the energy function of states of original images, fitted images, and the contour created, which are defined adequately. Optimization of the fitted images and the contour extracted is made with the gradient decrease method in relation to their states. The energy of the states of images is represented as a sum of a few terms corresponding to various effects of the method. The energy of states subjected to the minimization process determines the architecture and dynamics of the neural network.!6
展开▼