Vol. 28, Issue 2, pp. 81-94 (1998)

Abstract

The problem of extraction of a contour from highly noised images is discussed. The method for contour extraction investigated is known in the literature as regularisation with discontinuities. This method, when applied to highly noised images, gives good results and, moreover, may by adapted to a simple cellular neural network. This possibility allows a real-time contour extraction to be performed. The idea of the method of regularisation with discontinuities is presented and one of its variants for the case of temperature noise is used for illustration. The method discussed is examined by means of different functions of the extinction noise temperature.