In mineral processing, the liberation of valuable mineral is of key importance in achieving high recoveries from downstream separation processes such as froth flotation and gravity concentration. To quantify mineral liberation, information on ore texture of the parent rock as well as properties of comminuted particles is essential. These properties have been quantified by statistical measures such as the proximity function and covariance function, which were extracted from SEM images of parent rock and particle polished sections, using convenient and efficient image analysis techniques based on Labview (TM) software. To quantify fully liberated particles, a phase specific line segment function has been introduced and evaluated by placing random line segments on the image. It was also found that the ore texture assumptions made by Barbery are not valid for the high grade sulphide ore tested and the general applicability of these assumptions is therefore questionable. Using the measured information above, predictive liberation models to quantify volumetric grade distribution of particles in 1D and 3D have been developed based on Barbery's work. Results show that the grade distributions of composite particles predicted from the proposed 1D model is closest to measured data than those of Barbery's 1D model. The predictions using the proposed 3D model are similar to those predicted from Barbery's model and are considered more realistic as the model does not rely on assumed ore texture but on measurements made on the parent rock and particle sections. (C) 2013 Elsevier Ltd. All rights reserved.