ANALYSIS OF SKELETONIZATION METHODS FOR IRREGULAR TEXTURE IN UNIQUE IMAGE DESCRIPTORS SEARCHING PROBLEM

Abstract

The problem of texture analysis, identification and recognition was investigated. The object of study is the process of grayscale image
skeletonization. The subject of research is the iterative and non-iterative texture skeletonization methods for searching unique image
descriptors in the problem of authentication object. The purpose of the work is to select the appropriate method for inhomogeneous irregular texture skeletonization. A review of texture types and skeletonization algorithms classification, advantages and disadvantages of various types of algorithms was given. Computational experiments and comparison of iterative Zhang-Suen and Guo-Hall algorithms and non-iterative method of the Blum’s median axis transformation
that’s applied to the sample images inhomogeneous irregular paper textures with microfibers were made. The problem of the comparative
analysis of texture skeletonization algorithms for images of inhomogeneous irregular textures was solved. Was developed quality criterion which uses information about the number of skeleton texture branches, which depends on the number of boundary points of the skeleton. Software that implements the discussed skeletonization algorithms and allows to rank them according to selected criteria was developed. The experimental results allow us to recommend the algorithm of Zhang-Suen for practical use to solve the problem of image skeletonization of inhomogeneous irregular textures.