Improvement of building extraction using decision fusion of locally and globally enhanced IKONOS images

In this paper a fully automated algorithm for building extraction from remote sensing IKONOS images is presented. Local and global enhancement of an original image improves the rate of building detection in some cases. However, some undesirable effects could occur due to image enhancement. As a result the Bayesian classification method which has been previously used could result in errors. To deal with such problems, decision fusion is used together with a shadow-based verification step to achieve a better result from locally and globally enhanced classified images. Experimental results justify the efficiency of the proposed method in dealing with the problem of building extraction in IKONOS images.