Fourier transform‐based windowed adaptive switching minimum filter for reducing periodic noise from digital images

Abstract

This paper presents a windowed adaptive switching minimum filter in frequency domain to restore images corrupted by periodic noise. Periodic noise frequencies that spread throughout the spatial domain image concentrate in frequency domain image as star-shaped peak regions. The proposed algorithm incorporates distinct stages of noisy frequency detection and correction. The noisy frequency detection stage has peak detection and noise map generation sub-stages to effectively identify noisy peak areas into a binary flag image from the directional image of the origin shifted Fourier transformed corrupted image. The proposed noise correction scheme restores the detected noisy areas of the corrupted frequency domain image with the minimum of nearest possible uncorrupted frequencies. Finally, inverse shifting and inverse Fourier transform operations generates the restored image. Experimental results in terms of subjective and objective metrics demarcate that the proposed periodic noise reduction filter is more effective in restoring images corrupted with periodic noise than other filters used in the comparative study.