Detecting irregularities in regular patterns

This study compares three different methods designed for detecting irregularities from regular dot patterns. Frequency domain information is used to split an original regular pattern into two images: the first image contains the perfect repeating pattern and the second one includes all irregularities in the original image. The methods are based on the Fourier transform, but they differ in how they separate or utilize the regular and irregular image parts. Performances of these methods are compared, and their strengths and weaknesses are discussed.