A Bayesian/Monte Carlo segmentation method for images dominated by Gaussian noise

A description is given of a thresholding algorithm that rapidly separates foreground objects from background clutter in images whose dominant feature is zero-mean Gaussian noise. Such images have been found to occur in digital radiography applications in which manufactured parts are imaged by a solid-state camera. The motivation behind the algorithm is discussed in terms of the requirements of an imaging system for nearly-real-time radiography in an industrial environment. The individual steps of the process are described, and the robustness of the technique with respect to signal-to-noise ratio and with respect to object size is discussed