Method of detection similar elements in textures with irregular structure

Abstract

We consider the problem of repetitive elements recognition with heterogeneous image texture. The complexity of solving problems from this class is connected to obtaining precise solution that is sensitive to image with irregular texture. The accuracy depends on correctness of original element selection from which the searching process will start. To solve the problem we propose new synthetic approach that combines statistical methods and machine learning method that allows obtaining resistant and accurate solution. The results of this algorithm are used for detection of reptile skin structure.