A novel approach to area fraction measurement is developed to deal with the limitations of the segmentation method. The new approach takes advantage of the statistical nature of the noise within an image to deconvolute the image histogram and calculate, by means of fitting, the area fractions. Both segmentation and noise deconvolution approaches are critically discussed, and their limitations and advantages tested using two extreme backscattered electron images of a metal matrix composite. The deconvolution approach is shown to produce accurate results in a situation in which the segmentation approach fails.
