Adaptive noise cancelling applied to image restoration

The one-dimensional adaptive noise cancelling principle is applied to two-dimensional imagery to remove line-interferences and random noise. For destriping LANDSAT or TM images, the preliminary results indicate that application of the adaptive cancelling principle is a promising and effective approach. Two different implementations are discussed. One models the stripings as periodic interferences and includes such a model in the adaptive cancelling scheme. The other one adaptively ´zero-means´ the striped image on a line-by-line basis. For ´cancelling´ wide-band noise, a novel bootstrap scheme is used to overcome the unavailability of the secondary noise source required by the adaptive cancelling scheme. The result is satisfactory and this novel scheme is applicable to time-signals as well.