Fusion of IKONOS Remote Sensing Filtered Images Using Shadow Information to Improve the Rate of Building Extraction in Urban Images

Deficiency of unsharp mask filter in elimination of some regions in IKONOS remote sensing urban image is one of serious difficulties in building extraction from such images. Sometimes, saturation of intensity levels in filtered image makes some regions of image disappear. As a compensation for this issue, in this paper a method for fusion of unsharp mask filtered image and histogram equalized image is presented. In the first step, fusion of filtered images is accomplished. Since shadows give some information about the location of buildings, fusion of filtered images with considering the shadow location can be a satisfactory cure for elimination of image components such as buildings. Afterward, Bayesian classifier is applied to the fused laplacian and edge images to extract the buildings. Experimental results justify application of the proposed method in building extraction for IKONOS remote sensing images.