Straight edge extraction and localization on noisy images

The paper describes a novel method of localizing line segments and estimating their start and end points. The lines are ordered within the image by detecting the orientation of straight edges. Image processing structure is considered which includes directional gradient processing for obtaining straight lines from edges and filtering of lines with the prescribed line slope. It permits to order lines of this slope according to their gradient values and their lengths. Extraction of a gradient line profile is used for start and end points estimation by thresholding. The problem considered is related to reduction of an original grey-level image to a set of geometric edge position parameters. This set may be used for construction primary features for image understanding and storing. It is also useful for a comparison of different images of the same scene and for finding the mutual points for images matching.