The quality of the pharmaceutical powder mixing strongly influences the potency and properties of the tablets. Furthermore, in order to optimize powder mixing processes, fast and accurate methods are sought for determining the concentration of powder mixtures. This work describes an optical method based on an image processing system to determine concentration of powders in a mixture that can be sensed by its gray-level values.The methodology was developed utilizing univariate and multivariate gray image analysis. The effect of difference in mean gray value (DMGV) on the precision and accuracy of the vision system was determined. The results that were obtained from the vision system were compared to those obtained with near-infrared spectroscopy.The multivariate gray image analysis methodology using frequency of each gray-level value (Phi) of gray scale as the independent variables obtained the best predictions. The vision system produced high precision with a standard deviation of about 1.7% w/w and a 2.6% relative standard deviation. The accuracy indicated by the root mean standard error of prediction (RMSEP) was around 1.7% w/w. The mixtures analyzed with the vision system and the near-infrared spectrometer yielded similar results. Analysis of all results confirms that the vision system is an adequate technology for determination of powder mixture concentrations. (C) 2003 Elsevier B.V. All rights reserved.