A new method to quantify parameters of membrane morphology from electron microscopy micrographs by texture recognition Carles Torras a,b,n , Dom  enec Puig c , Miguel A ´ ngel Garcı ´a d a Bioenergy and Biofuels Division, Catalonia Institute for Energy Research (IREC), C/Marcel  lı ´ Domingo, 2, 43007 Tarragona, Catalunya, Spain b SYSTEMIC Research Group, Departament d’Enginyeria Quı ´mica, Universitat Rovira i Virgili, Av. Paı ¨sos Catalans, 26, 43007 Tarragona, Catalunya, Spain c Intelligent Robotics and Computer Vision Group, Department of Computer Science and Mathematics, Universitat Rovira i Virgili, Av. Paı ¨sos Catalans, 26, 43007 Tarragona, Catalunya, Spain d Department of Electronic and Communications Technology, Universidad Auto ´noma de Madrid, Francisco Tomas y Valiente, 11, Madrid, Spain article info Article history: Received 15 December 2010 Received in revised form 1 June 2011 Accepted 13 June 2011 Available online 24 June 2011 Keywords: Membranes Porous media Morphology Computation Texture recognition Imaging abstract A new method has been developed in order to automatically quantify parameters of membrane morphology from micrographs obtained through microscopy techniques. The parameters estimated by this algorithm are: pore size distribution, porosity, pore symmetry, regularity and tortuosity, as well as various statistical measures. These properties determine the performance of a membrane. The proposed method is based on texture recognition. It first identifies the pores present in the membrane from a cross-section micrograph of it, then labels them and finally makes the corresponding measurements. The main difference and advantage of this technique with respect to previous proposals is that the algorithm does not perform generic particle recognition, but direct scanning of typical pore structures and no user decisions are needed in all the steps of the process. Additionally, the proposed technique does not only determine typical parameters, such as pore size, but also particular characteristics of membrane topology, such as symmetry. The source information consists of cross-section membrane micrographs that can be typically obtained from electron microscopy (scanning or transmission), as well as from other types of microscopy, which are the most common acquisition techniques used by membranologists. The system provides quantitative, systematic and fast results, which represents a significant advance in the field of membrane analysis. & 2011 Elsevier Ltd. All rights reserved. 1. Introduction Membranes are selective barriers able to separate different types of chemical components from a mixture by following different principles mainly determined by the driven-force, which can be concentration, electrochemical, pressure, etc., trans-membrane gra- dient. Trans-membrane pressure gradient is one of the most used driven-forces applied to micro-, ultra- and nano-filtration, as well as to reverse osmosis. In the first two cases, the separation is based on the differences in the size of the compounds present in the mixture, while in the third one, diffusion phenomena plays an important role. Thus, one of the key factors of a membrane is the size and distribution of its pores. Therefore, porous membranes are typically used in these operations (Kools, 1998; Mulder, 1997). The performance of a membrane for micro-, ultra- and nano- filtration highly depends on its porous structure, which not only determines its separation capability but also its permeability. Membranes must be properly prepared and characterized in order to yield a good performance. There are two basic methods to characterize the morphology of membranes: direct and indirect measurements. Indirect measurements refer to experimental tests that allow one to correlate the performance of the mem- brane with its mean morphological characteristics (Palacios, 1999). Direct measurements refer to the visualization of the porous structure, which allows for a deep and detailed inspection of it (Torras Font, 2005). Due to the thickness of a membrane and the microscopic size of its pores, electron microscopy is mainly used, typically Scanning Electron Microscopy (SEM) or Transmis- sion Electron Microscopy (TEM). After preparing the sample properly, membrane micrographs can be obtained, from where only qualitative results can be obtained. Afterwards, although general algorithms can be used to obtain numerical results, they are too generic, not allowing the estimation of singular properties like membrane asymmetry (Image-Pro, 2009; King et al., 2002). Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/ces Chemical Engineering Science 0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.ces.2011.06.013 n Corresponding author at: Bioenergy and Biofuels Division, Catalonia Institute for Energy Research (IREC), C/Marcel  lı ´ Domingo, 2. 43007 Tarragona, Catalunya, Spain. Tel.: þ34 977 202 444. E-mail address: ctorras@irec.cat (C. Torras). Chemical Engineering Science 66 (2011) 4582–4594