Suppression of periodic and quasi-periodic patterns in the Fourier frequency domain

Abstract

The goal of this semester project is to create software which allows removing periodic and quasi-periodic pattern in images. Periodic noise in images or in image sequences can signify a serious problem in automatic image analysing example given in counting the number of cells in cell population or in the optical tracking in a Petri dish. Our software, a plug-in designed for the open source image editing and processing software ImageJ, provides algorithms to remove periodic structures in a manual, an automatic or in a semi-automatic way. The program allows also to process images sequences automatically. The software is doing the noise removal by analyzing the power spectrum of the Fourier transform of the image. Periodic and quasi-periodic patterns in images lead to peaks in the power spectrum. These peaks can be detected by different manners and they are attenuated or suppressed afterwards. The new power spectrum without the peaks is used to reconstruct the initial image free of the periodic structures.