Porous Media Characterization by Micro-Tomographic Image Processing

n this thesis, we have focused our attention on the characterization of porous media through micro-tomographic image processing. A porous medium can be simply seen as a solid material with "holes" in it, which, connected or isolated, may or may not eventually allow the flow of one or more fluids. They have various applications in common practice and are widely used in many disciplines, both scientific and industrial. Porous media are strongly characterized by their internal microstructure, which needs to be accurately described in order to determine their performance and macroscopic properties. Despite recent technological advances and the introduction of imaging techniques such as X-ray micro-tomography, methods to characterize quantitatively the porous media internal structure are still few and related to some specific applications. To this regards, we propose new algorithms for the analysis of the 3D micro-architecture of porous media based on image processing and the mathematical morphology theory. In particular, the opening operator properties have been exploited in the "successive openings" algorithm which represents the starting point for the construction of three morphological synthetic indicators. They are the dimensional curves, Pore Size Distribution (PoSD) and Trabecular Size Distribution (TrSD), which provide information about the pores or solid phase structure, the connectivity curves, which allow to identify how the structural elements are interconnected, and the effective porosity, which represents the porous fraction concerned with the transport of fluids. Experimental results show that the proposed indicators together represent an effective tool for the porous media internal structure characterization Since noise is a primary cause of reduced image analysis capability in micro-CT, we have dedicated a part of our research to the reduction of the strong noise that corrupts tomographic images. After evaluating experimentally the characteristics of noise, we propose a filtering technique for correlate noise based on the Block-Matching 3D (BM3D) algorithm. Experimental results prove the proposed technique effectiveness and its potential to improve the performance of the algorithms proposed in the first part. Although micro-tomographic image processing presents considerable difficulties, both for the intrinsic characteristics of the images, and for the nature of analyzed objects, this thesis proves that reliable and useful indications about the structure of porous media can be obtained through the use of the mathematical morphology theory.