The textural analysis of gravity data using co-occurrence matrices

Abstract

Images derived from measurements of the strength of the Earth's gravity field are made routinely and are used for many purposes, such as mineral exploration. Textural analysis of these datasets using grey-level co-ocurrence matrices (GLCMs) is a useful method for enhancing subtle detail, and are frequently used as an aid to interpretation. The GLCM textural measures involve a vector that connects pairs of pixels within a kernel that is moved over the image. This paper introduces GLCMs which use vectors designed specifically for the circular features that are associated with short wavelength anomalies in the Earth's gravitational field. The GLCM vector, instead of being the same at each point within the kernel, is made to follow the contours of constant field value of simple gravity anomalies. The use of a textural measure such as the inverse difference moment, which attains a maximum value when all the image pixels that are compared have the same value, then yields a strong response at the central locations of the features of interest. The filters are demonstrated both on synthetic data and on gravity data from South Africa.