Extracting Linear Features from Images Using Pyramids

A method is described of extracting linear features from images. The approach is to construct a series of lower-resolution versions of the original image (a pyramid), and to look for lines in these images. A line in a low-resolution image corresponds to a thicker linear feature in a high-resolution image. The position and extent of this linear feature is calculated from the low-resolution image, and a threshold is found which, when applied in the neighborhood of the feature in the high-resolution image, segments the linear feature from its background. Advantages of the method are that only the parts of the image in the neighborhood of linear features need be thresholded, and that different thresholds may be used to extract the various linear features in the image. The support of the Defense Advanced Research Projects Agency and the U.S. Army Night Vision Laboratory under Contract DAAG-53-76C-0138 (DARPA Order 3206) is gratefully acknowledged, as is the help of Kathryn Riley in preparing this paper. The author is Iso grateful to Shmuel Peleg and Les Kitchen who supplied the line-enhancement programs.