Linearization of mammograms using parameters derived from noise characteristics

A method is proposed for converting digitized mammograms to a normalized representation, in which pixel values are linearly related to the logarithm of x-ray exposure. This method is based on a signal dependent noise model. By exploiting this model unknown parameters of the non-linear response of the film-screen system can be estimated from signal dependence of the image noise. The method was applied to a series of 1372 mammograms acquired over a period of 8 years. Sudden changes in estimated parameters corresponded well with the introduction of new film-screen systems.