Detection of quasiperiodic textures with the use of two-dimensional power spectrum characteristics

Abstract

An algorithm for detecting quasiperiodic textures is proposed, which is based on specific features of two-dimensional spectra obtained by using the discrete Fourier transform. The power spectra of the fragments are analyzed independently of each other, which provides the computational effectiveness of the detection algorithm. Examples demonstrating the use of the detection algorithm in the problem of image segmentation into two texture classes (periodic and nonperiodic) are presented. It is shown that the algorithm stably operates with significant deformations in the lattice of periodic textures.