Although shadows are considered useful for the identification of objects that rise up from the ground, they may made the extraction of information difficult when they hide lower objects. In this paper we introduce the conceptual development and show examples of the application of a method for the segmentation of shadows in satellite and aerial imagery that is based on the analysis of the co-occurrence matrix. The co-occurrence matrix shows the relationship between a pixel and its neighborhood, which allows to identify dark and uniform areas that correspond to shadows. Taking this in mind, a method is proposed to analyse the main diagonal of the co-occurrence matrix and to identify the cluster that is related to shadows. The mentioned segmentation enables to obtain the threshold along the main diagonal. A contextual function can be derived from the best perpendicular line to the main diagonal of the co-occurrence matrix that crossed the diagonal at the threshold point. Tests with aerial and satellite images are presented to illustrate the method. It is shown that the method is adaptive. It can be used in 8 bits images as well as extended to higher radiometric resolutions and it is not affected by the atmospheric contribution.
Image Processing; Remote Sensing; Segmentation; Shadows
