Adaptive threshold based frequency domain filter for periodic noise reduction

Restoration of images corrupted with periodic noise is a fundamental task in digital image processing since periodic noise affects all imaging processes. The paper presents an adaptive threshold based frequency domain filter for denoising periodic noise from corrupted digital images. The algorithm applies forward origin shifting and Fourier transform to convert spatial domain image to frequency domain image. The proposed algorithm adaptively determines the threshold function for identifying the noisy peak areas of frequency domain image. The identified noisy frequencies corresponding to noisy peak areas of frequency domain image are diffused by the minimum filter for the effective restoration of frequency domain images. The restored frequency domain image is applied with inverses of Fourier transform and shifting operations to reconstruct the final restored image in spatial domain. Experimental analysis on different images with varying noisy conditions shows that the proposed algorithm is capable of producing better restored images than other algorithms used in the comparative study.