Suppression of periodic interference from images via transform domain processing

The removal of periodic interference in computer images is a relatively well-understood problem. Assuming no a priori knowledge of the periodic signal, one may apply a 2D-DFT to an image to yield the spectral composition. When the periodic signal contains sufficient magnitude, spectral peaks can be identified in the spectral domain and may be removed by replacing the spectral peaks with an average or median value of the surrounding components. Application of a 2D-IDFT will result in a restored image. The process described is adequate for periodic signals that introduce an integer number of cycles throughout the image, however, when a non-integer number of cycles exists, the technique fails to remove a sufficient amount of corrupting signal energy. A robust algorithm has been developed to eliminate the manual process of identifying and removing the periodic interference. The Enhanced Median Adjusted Detection (EMAD) algorithm examines the values surrounding a spectral component and adaptively determines if the spectral component is a spectral peak. Once identified as a peak, application of a median filter suppresses the peak.