The intersective Hough transform for geophysical applications

Abstract

The Hough transform is a widely used tool for alignment identification, but the voting scheme usually used in this transform presents serious problems when alignments are not exact. We propose a voting scheme based directly on intersections in Hough space, which is more efficient and solves the above mentioned problems; the proposed scheme also provides a straightforward way of quantifying each alignment. The intersective approach can even dispense with gridded Hough spaces and accumulator cells, through direct clustering of intersections, so that the whole alignment identification process can be done automatically. A synthetic example representing a geophysical application of the Hough transform, identification of inexact alignments in the presence of noise, is used to illustrate the proposed method and to discuss further applications. An optional data-weighting scheme takes into account possible uncertainties in the data. For both exact and inexact alignments, the intersective method yields better results than the traditional one.