Computed tomography (CT) is a powerful technique that can be used to image multi-phase flow in three dimensions. In this study, X-ray CT is used to image trickle flow in a stationary packed bed with a spatial resolution of 118 mu m/pixel. Errors introduced during the radiograph acquisition and tomographic reconstruction of the volume image make it difficult to identify the three phases (gas, liquid and solid) and in particular the interfacial areas. A novel post-processing strategy based on the matrix convolution operation and a priori knowledge of the shape of the particles is developed that makes it possible to accurately identify the phase interfaces in an unbiased way. The result is a ternary three-dimensional image where each voxel is one of gas, liquid or solid. From this, the gas-liquid, gas-solid and liquid-solid interfacial areas can be calculated. The proposed procedure yields images that are superior to those obtained from the usually employed thresholding operation. (C) 2007 Elsevier Ltd. All rights reserved.