This paper presents a methodology based on spatial autocorrelation function for assessing homogeneity of a mixture and more particularly for determining the size and the number of defects in the bulk of a mixture or materials. Intensity and scale of segregation are used as conceptual tools for assessing mixture homogeneity, and have been specially adapted for this particular case. Their performances are investigated for detecting defects in a product manufactured from a dry powder mixture, a bipolar plate used in fuel cell technology, through a simulation of the product on Matlab (92). (C) 2010 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.