Image filtering based on discrete cosine transform using robust indicators of non-Gaussian distributions

N. A. Dementev, A. A. Butov

Ulyanovsk State University, Russia

Abstract: The paper considers a filter for images based on a discrete cosine transform using a robust indicator of non-Gaussian distributions. An algorithm is presented that varies its parameters depending on the area under study, decreasing the threshold value in the vicinity of the boundaries of objects. The proposed method allows you to effectively deal with noise in the image and preserve the boundaries of objects.

Keywords: discrete cosine transform, robust kurtosis estimation, image filtering.

UDC: 519.87 + 004.9 + 573.2

Received: 19.12.2022
Revised: 07.02.2023