Use of homomorphic transforms for low contrast edge/detail detection and filtering of MM-band images

MM-band radar and radiometric images are usually corrupted by data-dependent noise (with Gaussian distribution) being, for instance, the mixture of additive and multiplicative components. This causes problems in direct application of many filtering algorithms specially designed for the cases of dominant influence of additive or multiplicative noise, but not for intermediate situations. Moreover, many edge/detail detectors that do not take the aforementioned noise peculiarities into account also fail to perform well. Besides, if the multiplicative noise is a prevailing factor then the traditional detail detectors indicate the details having equal but positive or negative contrast in respect to surrounding background with different reliability. This prevents the appropriate preservation of negative contrast details when nonlinear in particular, adaptive filters are used for image processing. So here we consider the use of homomorphic transforms in order to avoid these problems. The transform should be selected in such a way that after its application the noise becomes additive and its variance remains constant for all regions of the image