We introduce a computational analysis workflow to access internal and external properties of solid objects using non-destructive imaging techniques that rely on X-ray imaging. The goal is to process and quantify structures from material science sample cross-sections. The algorithms can differentiate porous media (high density material) from void (background, low density media) using a Boolean classifier, and extract features such as volume, surface area, granularity spectrum, porosity, among others. Our workflow, Quantifying Computed Tomographic Images (Quant-CT), leverages several algorithms from ImageJ such as fast statistical region merging and the 3D object counter, which includes schemes for embarrassing parallelization of stacks, bilateral filtering using a 3D kernel, and proposes an algorithm to handle over-segmentation using statistical moments. Quant-CT supports quick user interaction, including the ability for the user to train the algorithm via subsamples to provide its core algorithms with automated parameterization. Quant-CT plugin is currently available for testing to personnel from the Advanced Light Source and Earth Sciences Divisions and Energy Frontier Research Center (EFRC), LBNL, as part of research on porous materials used for understanding the processes in fluid-rock systems for geologic sequestration of CO2, and development of technology for the safe storage of CO2 in deep subsurface rock formations. We overview the material science problem, describe our implementation, and demonstrate our plugin on micro computed tomography images. This paper targets end-users, with relevant information for developers to extend current capabilities.
segmentation, feature extraction, material science, carbon sequestration
Dani Ushizima's work focuses on image analysis and pattern recognition applied to diverse scientific domains - images range from biomedical pictures to porous material, e.g. micro-tomography of materials with applications to carbon sequestration. She has acted as Principal Investigator of the project Quantitative Image Analysis for Computational Modeling (LDRD-DOE) and co-P.I. of the projects: Visualization and Analysis for Nanoscale Control of Geologic Carbon Dioxide (Scidac-e) and From images to models to computational input (LDRD-DOE). Interests include computer vision, machine learning, signal processing, quantitative microscopy, and high-performance computing.
Presenting author: Daniela M. Ushizima
Organisation: Lawrence Berkeley National Laboratory