Segmentation of grey scale sampled images with bimodal source models

A gray-level-sampled image of a bimodal source will have some deviations from a perfect bimodal distribution due to added source, partial volume effects, and slow variations in source intensity. If the high contrast source image is to be accurately estimated from the sampled image, the characteristics of the image acquisition system must be selected to preserve the required information. These characteristics include the related effects of spatial sampling frequency, number of bits per pixel, aperture functions, and measurement noise. A method for estimating the modes of the underlying bimodal distribution is presented, and the results are used in algorithms which recover a bimodal source description. Examples demonstrate these effects.<
>