Texture analysis of terrain images:exploiting directional properties of a local-feature statistics operator using small masks

Telfer, Duncan J. and Wilson, Lionel (1996) Texture analysis of terrain images:exploiting directional properties of a local-feature statistics operator using small masks. In: Proceedings of SPIE - The International Society for Optical Engineering. Proceedings of SPIE - The International Society for Optical Engineering . UNSPECIFIED, Orlando, FL, USA, pp. 148-158. ISBN 0819421340

Abstract

The texture discrimination properties of a directional operator, using a window based histogram correlation technique, are described for test images and natural scene imagery. The natural image examples are drawn from digitized monochrome aerial photographs of volcanic terrain, and of coastal dune fields in the region of West Lancashire, England. The operator masks use directionally encoded local difference parameters and are of two sizes, 3 by 3 and 2 by 2 pixels, scanning over a window size of 25 by 25 pixels to create an autocorrelation map, normalized to the output gray-scale range. Comparisons of the outputs are discussed in terms of ability to detect specific types of texture feature associated with known lava flow regimes demonstrate the viability of the technique down to single pixel resolution, particularly for folded or striated topography. The vegetation cover of the volcanic terrain is also characterized in terms of its response to these texture operators. The output from the dune field images, which contain a wider variety of vegetation cover, also demonstrates the discrimination capability of the technique and its potential usefulness in environmental and conservation management.