Shadow detection is a crucial task in high-resolution remote-sensing image\nprocessing. Various shadow detection methods have been explored during\nthe last decades. These methods did improve the detection accuracy but are\nstill not robust enough to get satisfactory results for failing to extract enough\ninformation from the original images. To take full advantage of various features\nof shadows, a new method combining edges information with the spectral\nand spatial information is proposed in this paper. As known, edge is one\nof the most important characteristics in the high-resolution remote-sensing\nimages. Unfortunately, in shadow detection, it is a high-risk strategy to determine\nwhether a pixel is the edge or not strictly because intensity values on\nshadow boundaries are always between those in shadow and non-shadow\nareas. Therefore, a soft edge description model is developed to describe the\ndegree of each pixel belonging to the edges or not. Sequentially, the soft edge\ndescription is incorporating to a fuzzy clustering procedure based on HMRF\n(Hidden Markov Random Fields), in which more appropriate spatial contextual\ninformation can be used. More concretely, it consists of two components:\nthe soft edge description model and an iterative shadow detection algorithm.\nExperiments on several remote sensing images have shown that the proposed\nmethod can obtain more accurate shadow detection results.
Loading....