About One Method of Differentiation of a Flat Discrete Planar Curve in Image Processing

The problem of receiving points with high curvature (singular points) of contours for identification of the shape of objects on images is solved. Analysis of existing methods of numerical differentiation in the given aspect is held. The new method of differentiation of the flat discretely defined curves, which are dots (pixels) of circuits, based on variations of Arch Height method is considered. Features of such method of differentiation are shown using various formulas of calculation of a derivative. Dependency aspects of the accuracy of the derivative on the chord length are analyzed. It is shown, that with an increase in its length differentiation accuracy degrades, and the result tends to the module of curvature of a curve at the given point. Comparison of the developed method with other known methods is made. The analysis of area of applicability and variability of parameters of differentiation is made. The accuracy aspects of calculation of derivatives for various parameters of differentiation are investigated. Examples of differentiation of various curves, both set analytically, and the functions-contours received from real images are considered. It is shown, that the offered method allows to get rid of the ambiguity in position of points of a contour with high curvature and consequently to raise quality of recognition of the shape of objects. Possible scopes of the given method in various areas of science and technics are stated.

image processing, pattern recognition, computer geometry, differentiation, singular points, analysis of contour