Formation of the Image Topological Characteristics Based on Two-Dimensional Variations and Their Application for Object and Noise Detection

Abstract

A variation-based approach to formation of the topological characteristics of two-dimensional signals is proposed. Application of these characteristics for the metric estimation of the local parameters of images is considered. Concepts of the object convexity size and amplitude indices are introduced and a technique for their calculation is proposed. The use of these parameters for detecting the noise and the objects of different dimensions in grayscale and multispectral images is demonstrated. Theoretical conclusions are confirmed experimentally.