Gabor-based spectral domain automated notch-reject filter for quasi-periodic noise reduction from digital images

Well-organized restoration techniques for attenuating the impact of periodic/quasi-periodic noise structures from digital images is one of the significant research fields in modern days. These are encountered in various imagery applications like remote-sensing (satellite, aerial), the digitization of canvas paintings, etc. In this paper, a novel spectral domain algorithm for periodic/quasi-periodic de-noising has been presented where an Automated Notch-Reject Filter (ANRF) is lucratively used to remove unwanted periodic patterns from Gabor-transformed corrupted images. As an initial stage, the Low-Frequency Region (LFR) has been conserved ingeniously by finding squared spectral difference after representing the image spectrum as multiple populations. Thereafter, the contrast of any corrupted image spectrum has been increased using Gabor transform for making the noisy components more prominent. Then, an adaptive exponential thresholding procedure has been applied efficiently for detecting those noisy components. The final stage of our proposed algorithm is to filter out those noisy components properly where a novel adaptive notch-reject filter has been applied along with an automated control of filtering profile in proportion to different noise spectrum profile. The supremacy of our algorithm over other state-of-the-art algorithms has been productively established with the help of experimental results in terms of visual and statistical metrics.