Thinning is an iterative process which erodes an object layer by layer until only a 'skeleton' or 'medial axis' of the object remains. A medial axis with high quality lays the foundation for quantifying a porous structure under a pore scale, which has a major effect on many engineering processes and applications. In this study, an algorithm was developed to compute the medial axis and radius of a pore in porous medium simultaneously from 3D images by conducting a labelling process. This ensures that every point possesses a number denoting its distance to the nearest solid. We used 231 computer-generated models with typically ordered porous structures and 3D images from an active carbon system acquired using X-ray computed tomography to validate the algorithm. By comparing the calculated porous volume with the actual porous volume of 30 computer-generated single pore models, we determined that the relative error of the porous volume was <6%. In addition, the radius of a pore in the reconstructed model of the active carbon system, which was measured manually using Imagej, an image processing software, and calculated automatically by the proposed algorithm, shared the same value. The results indicate that the proposed algorithm can successfully compute the medial axis and radius of the void space for any porous system, irrespective of the shape of solid particles or the geometry of the pores. (C) 2018 Elsevier B.V. All rights reserved.