A DCT based method and system for reducing noise from an image is described. First, we perform nosie modeling and estimation of the imaging source using flat fields to obtain a set of masks and LUTs that characterize the souce of noise. Second, the image is represented in a pyramid structure by recursively using the DCT as a filter bank on the original image and then on the low frequency hand. This representation is complete, reversible and allows us to process the image at different resolutions. Third, a Wiener variant filter using the estimated masks and LUTs is used to perform nosie reduction on the image in the pyramid representation. Finally, the image is reconstructed by recursively using the IDCT as a filter bank in the inverse direction starting at the highest level of the pyramid.
展开▼