Shadow Detection Using Object Oriented Segmentation, Its Analysis And Removal From High Resolution Remote Sensing Images

Very high resolution satellite images available from satellites such as QuickBird, IKONOS etc are usually affected with shadows. These shadows reduces the information content of the images. In this paper, an object oriented shadow detection method is used to detect the shadows. In this method each object in the input image is extracted through a segmentation process. Suspected shadows are detected by comparing the threshold obtained from the original image and the gray scale average of each object in the segmented image. Dark objects which may be included in the suspected shadows are eliminated by com-paring the gray scale average in different image bands. For the restoration of information, relative radiometric correction is used. For this inner-outer outline lines along the shadow boundary are generated. For shadow removal, relative radiometric correction is applied to the homogeneous points obtained from IOOPL similarity matching. Experimental result shows that the proposed algorithm effectively detected majority of shadows within the high resolution satellite images compared to pixel based methods.