Automatic shadow detection and radiometric restoration on digital aerial images

This paper develops automatic procedures for shadow detection and radiometric restoration of features under cast shadows in aerial images. Techniques including recursive quad-tree image partition, adaptive thresholding, region growing segmentation, mathematical morphology, and local histogram matching are applied to automatically extract the regions of building-cast shadows and to restore the brightness of those pixels within the shadow regions. Empirical results show that the detail information of features within the shadow regions is enhanced by applying the techniques developed in this paper.