Artefacts of Wavelet Filtration of Images and Their Elimination

Abstract—Wavelet filtration algorithms are widely used for image processing; the majority of them are of threshold character: the wavelet decomposition coefficient whose value is smaller than some threshold value is equated to zero, otherwise, this coefficient is subjected to some (mostly nonlinear) transform which is defined by the used threshold function. When filtering contrast images, such treatment of decomposition coefficients may lead to the so-called artefacts in the filtered image — in the neighborhood of the stepwise change in the amplitude of the filtered image, the pulses appear that are absent in the initial image, which significantly worsens the filtration quality. To eliminate such artefacts, several approaches are considered in this study and a method is proposed whose gist consists in posttreatment of the result of wavelet filtration of images by means of nonlinear spatial filtering algorithms. The calculation experiment performed in this study has made it possible to come to the conclusion that the proposed method is efficient to eliminate the artefacts of wavelet filtration of contrast images.