《Level Set Segmentation of Remotely Sensed》.pdf

Level Set Segmentation of Remotely Sensed Hyperspectral Images J. E. Ball, Student Member, IEEE and L.M. Bruce, Senior Member, IEEE Department of Electrical and Computer Engineering Mississippi State University Starkville, MS, USA jball@ Abstract—We present a semi-automated supervised hyper- applications of level sets in multispectral or hyperspectral remote spectral image segmentation algorithm based on the level set sensing. Keaton and Brokish used level sets to segment roads in pan methodology. In the proposed procedure, seed pixels are sharpened IKONOS multispectral images [9], DellAcqua, Gamba automatically selected by their similarity to the training and Prevedini extracted and tracked moving clouds in GOES and signatures, and speed functions that control the level set Meteosat IR satellite image sequences [10], and Harper and Reilly propagation are created based on pixel similarity to the seed used level sets to segment faces in RGB video [11]. signature and class discriminator functions. Two sub images In contrast to the few examples in remote sensing, level set methodologies have been used extensively and successfully in many from a remotely sensed HYDICE hyperspectral image of the