Measuring image structures using a multiscale orientation field
J.M. Coggins · 2002
A novel method for representing image orientation structure is used to measure the orientations of line segments in a series of increasingly blurred images. An algorithm for mapping filtered image data into an orientation feature space is defined. The algorithm is applied using four sets of filters. The results show that the algorithm effectively exploits redundancy in the feature values to yield robust inferences across a broad range of scales and through large amounts of blurring.>