Computation of grain size distribution in 2-D and 3-D binary images

Abstract

Grain Size Distribution is one of the basic measurements for sediment classification. The conventional methods for grain size distribution include the sieve method, the laser diffraction method, and the point-count method. We aimed to develop a robust computer code that simulates these conventional methods. The code can measure grain size distribution on 2-D and 3-D binary images using a watershed algorithm to extract out individual grains, and using principal component algorithms to find the principal axes. The outputs include grain radius for different principal axes, grain volume, grain surface area, principal axes inclinations and azimuths, and the number of contacts for each grain. The calculated distribution can be volume-based, frequency-based, or grid-based. Digital microstructures used in this study include (1) identical sphere packs including a simple cubic pack and a Finney pack, and (2) natural rock geometry such as Berea sandstone, Castlegate sandstone, and Fontainebleau sandstone. Furthermore, we employed this code to provide additional value of information from μXCT images by using μXCT to create 2-D to 3-D model of the grain size distribution, solving what is commonly known as Wicksell's corpuscle problem. We showed that our workflow successfully models a generalized 2-D to 3-D grain size distribution for a particular set of natural rocks that we include in our study. We hope to be able to obtain more μXCT images in the future in order to create a universal model covering most types of natural rocks.