Estimation of grain-size distributions and associated parameters from digital images of sediment

A new technique to estimate the grain-size distribution (GSD) from a digital image of sediment is proposed, advancing the applicability of a suite of sedimentary ‘look-up-catalogue’ approaches originated by Rubin [Rubin, D.M., 2004. A simple autocorrelation algorithm for determining grain size from digital images of sediment. Journal of Sedimentary Research, 74(1): 160–165]. The outputs of an automated procedure to estimate the GSD from digital images of sediment are examined with reference to the distributions obtained from both manually sieving the corresponding sediment samples, and axial measurements made on the grains in the images. Measures of grain-size obtained from the imaging procedure correlate very well with grain-size measures derived from both the number-frequency and mass-frequency curve. As expected the GSD obtained from the new automated approach, based on kernel density, compared better with point counts because of a shared two-rather than three-dimensionality. The GSD shape is not always mimicked exactly, however the percentiles obtained from the cumulative GSD compare well with those from sieved distributions, which allow for the first time computation of graphical sorting and skewness from digital images of sediment which are accurate reflections of those measures obtained for point-count and sieved samples. The new approach allows for realistic GSDs from which the residual can be computed, thus a numerical criterion upon which the grain-size distribution from an image can be accepted or rejected. Finally, a method is presented whereby two-dimensional autocorrelogram fields are derived from image power spectra. Ellipsoids are fitted to isolines of autocorrelation coefficients, and from this the dominant axial lengths and orientation of these isolines which could facilitate computation of major and minor axis lengths of grains in sample images, as well as their dominant orientation. In turn, this could allow for parameterisation of axial ratios for parameterisation of 2D shape. Such analysis is completely automated, rapid and non-intrusive.