The antialiased textural analysis of aeromagnetic data $ G.R.J. Cooper School of Geosciences, University of the Witwatersrand, Johannesburg 2050, South Africa article info Article history: Received 13 September 2007 Received in revised form 28 March 2008 Accepted 11 April 2008 Keywords: Grey-level co-occurrence matrices (GLCM) Textural analysis abstract Textural analysis is a powerful tool that can be used to enhance subtle detail in raster datasets. It is often used as an input to data classification methods to improve the accuracy of the result. This paper suggests the application of unweighted area sampling to the calculation of texture using grey-level co-occurrence matrices because it gave much improved results compared to the standard method, particularly when global classes are used. Source code in Matlab format can be downloaded from the IAMG server at www.iamg.org/CGEditor/index.htm. & 2008 Elsevier Ltd. All rights reserved. 1. Introduction The texture of a grey-scale image is related to the local spatial distribution of the pixel values. Textural analysis has been applied to a wide range of Earth science datasets, such as potential field data (Dentith, 1995; Cooper, 2004; Cooper and Cowan, 2005), synthetic aperture radar data (Soh and Tsatsoulis, 1999), Landsat (Welch et al., 1990) and SPOT imagery (Marceau et al., 1990). One of the most common techniques of measuring texture is the use of grey-level co-occurrence matrices (GLCMs) (Haralick et al., 1973). GLCMs examine the relationship between pairs of pixels in a window which is moved through the data, measuring the probability that a pixel whose value lies within a given range is associated with a second pixel whose value lies within another given range. Once the GLCM matrix has been calculated (for a given window) then the next step is to calculate the texture from it. Although there are a large number of different types of textural measures only one will be discussed here since the antialiasing method described below can be applied to any of them. Inverse difference moments (IDM) are defined as (Haralick et al., 1973) IDM ¼ 1 N X m i¼1 X m j¼1 1 1 þði  jÞ 2 glcmði; jÞ (1) where glcm is the GLCM matrix, m is the number of grey levels (see below), and N is the total number of co- occurring data pairs. This has a maximum value, when all pixels within the window have the same value, and hence has a large value when the data are relatively smooth. Fig. 1 shows a simple synthetic magnetic anomaly and its IDM texture computed using data values one unit apart in the East–West direction. Anisotropic data will have a different texture in different directions and if a direction- independent textural measure is required, it can be obtained from the average of the texture calculated in several different directions. Although some work has been performed to show the value of the textural analysis of potential field data to complement standard geophysical filters, such as the vertical derivative (Dhu et al., 2000; Dentith et al., 2000), textural analysis has most commonly been used to improve the accuracy of classification and has been used in geophysical projects for this reason (Cole and Stettler, 2001; Holden et al., 2007). 2. Texture and class size As discussed in the introduction, the GLCM matrix examines the statistical relationship between pairs of Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/cageo Computers & Geosciences ARTICLE IN PRESS 0098-3004/$ - see front matter & 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.cageo.2008.04.009 $ Code available from server at http://www.iamg.org/CGEditor/ index.htm. E-mail addresses: gordon.cooper@geosciences.wits.ac.za, cooperg@geosciences.wits.ac.za (G.R.J. Cooper). URL: http://www.wits.ac.za/science/geophysics/gc.htm (G.R.J. Cooper). Computers & Geosciences 35 (2009) 586–591