A method of grey-level histogram threshold selection using fuzzy probability is reported. An image is treated as a set composed of two fuzzy subsets: ‘object’ and ‘background’. The grades of membership of grey levels in the ‘object’ subset are found with an experimentally determined membership function. The fuzzy probability that the ‘object’ points appear in an image is calculated from the membership function, and the frequency of occurrence of each grey level. This probability is equal to the area fraction of the objects in the image. Once the area fraction has been determined, the appropriate threshold can be selected. This method mimics the human process of locating boundaries of objects, and is especially useful for quantitative image analysis. It produces consistent thresholding with different users even when an image's histogram contains a broad. flat valley between its peaks.