Optimum Image Resolution of a micro-CT image to characterize shape descriptors of unconsolidated sand

Abstract

Resolution plays a fundamental role in any quantitative image analysis. Higher resolution images contain more details for further analysis but trade-off encounters when resultant smaller sample size raised question on representativeness of the whole sample. Image acquisition time and cost are also major issues that high resolution images have to face. To identify maximum image resolution that can avoid these issues as well as can provide accurate results in calculating shape factors, we study images of quartz sand grains acquired with four different resolutions. We present a comprehensive processing technique that can effectively extract individual grains from a 3D micro-CT image. Then we calculate equivalent diameter, volume and surface area of the grains at different resolutions. For all four resolutions, volume of grains shows very little change in lower two resolutions and almost no change through higher three resolutions, minimum of which can be considered as optimum. On the other hand, surface area for all the grains shows increasing trend with increasing resolution, but different in gradient. This different increasing trend can be explained by the surface rugosities and whether the image resolution can be able to resolve those. The higher two resolutions can effectively resolve surface irregularities of most of the grains, which is evident by their similar values of calculated surface areas. Therefore, minimum of these two resolutions can be considered as optimum image resolution in calculating shape factors for the studied grains.