Automatic orientation mapping of some types of soil fabric

Abstract A technique is described which can automatically delineate regions of micrographs of soils which have a similarly oriented microfabric. This is achieved by an initial pass over a digitized image using an intensity gradient algorithm to specify the orientation at each pixel. Each pixel is then given a value according to this computed direction. A second pass over the image examines the values attributed to each pixel. If, within the neighbourhood of a pixel, one particular general orientation class is dominant then the central pixel is assigned a grey-scale value appropriate to that class. The resulting image shows regions of the image which are aligned predominantly in a given direction. The resulting domain-segmented image may also be combined with the original to generate an image showing the normal grey tonal range, but overlaid with colour washes according to the general orientation of the features in the image at that location. Quantitative parameters, such as the degree and direction of preferred orientation, may be derived from both the initial intensity gradient analysis and also the domain-segmented image. All the steps may be combined in a batch processing mode to automatically analyse several images sequentially.