Powder mixing is one of the most impacting steps in the handling of pharmaceutical powders for tablet production due to its strong effect in the proper-ties and potency of the tablets. The monitoring of this process to detect the required mixing end point so that adequate properties and potency on the tablets are achieved requires continuous analysis of the powder concentration. However, its complex dynamics requires fast and accurate methods to determine the concentrations. Image analysis has already been shown to be suitable for this application. The objectives of this work were to further demonstrate its applicability to bright as well as dark powders, to determine the discrimination limit of the used vision system between powder's Mean Gray Values (MGV) in mixtures, and to develop a mathematical model to accurately predict calibration curves with minimum adjustable parameters. The study was emphasized for bright powders iti the region of MGV between 255 and 140 and for dark powders in the region between 139 and 0. Mixtures of powders with difference in density greater than 0.3 g/cm(3) showed considerably deviation in the prediction of concentrations. The discrimination limit between MGV attained was between 21 and 17 for both bright and dark powders. Below this range, unacceptable residuals larger than 8% w/w were obtained. (c) 2007 Elsevier B.V. All rights reserved.