Building Extraction Based on Hue Cluster Analysis in Complex Scene

Abstract

This paper presents an approach to extract building region in an image with complex scene. The algorithm highlights the edge characteristics of the target in image pre-processing step, uses the improved Hough transform line detection algorithm and the line optimization algorithm to keep the effective straight-line segments. Then it clusters the dominant hues of each effective segment's surrounding area, which will be made statistical analysis to group the straight-lines with building hue and keep the building lines with consistent hue. Finally, the building region is gotten in original image by the effective lines. Our experiments on dozens of pictures by this method and two other building extraction methods, demonstrate that this method is not only good for extracting the covered buildings in complex scene images, but also good for locating the buildings submerged in the background accurately. Keywords-building extraction; improved Hough transform; lines optization; hue clustering