Abstract
This paper introduces a new supervised technique to segment hyperspectral images: the Bayesian segmentation based on discriminative classification and on multilevel logistic (MLL) spatial prior. The approach is Bayesian and exploits both spectral and spatial information. Given a spectral vector, the posterior class probability distribution is modeled using multinomial logistic regression (MLR) which, being a discriminative model, allows to learn directly the boundaries between the decision regions and, thus, to successfully deal with high-dimensionality data. To control the machine complexity and, thus, its generalization capacity, the prior on the multinomial logistic vector is assumed to follow a componentwise independent Laplacian density. The vector of weights is computed via the fast sparse multinomial logistic regression (FSMLR), a variation of the sparse multinomial logistic regression (SMLR), conceived to deal with large data sets beyond the reach of the SMLR. To avoid the high computational complexity involved in estimating the Laplacian regularization parameter, we have also considered the Jeffreys prior, as it does not depend on any hyperparameter. The prior probability distribution on the class-label image is an MLL Markov-Gibbs distribution, which promotes segmentation results with equal neighboring class labels. The a-expansion optimization algorithm, a powerful graph-cut-based integer optimization tool, is used to compute the maximum a posteriori segmentation. The effectiveness of the proposed methodology is illustrated by comparing its performance with the state-of-the-art methods on synthetic and real hyperspectral image data sets. The reported results give clear evidence of the relevance of using both spatial and spectral information in hyperspectral image segmentation.

Abstract

Abstract

Abstract
There is a considerable interest in the development of automatic image analysis systems for dermoscopic images. The standard approach usually consists of three stages: (i) image segmentation, (ii) feature extraction and selection, and (iii) lesion classification. This paper evaluates the potential of an alternative approach, based on the Menzies method. It consists on the identification of the presence of 1 or more of 6 possible color classes, indicating that the lesion should be considered a potential melanoma. The Jeffries-Matusita (JM) and Transformed Divergence (TD) separability measures were used for an experimental evaluation with 28 dermoscopic images. In the most challenging case tested, with training identified in multiple images, 8 out of 15 class pairs were found to be well separable, or 13+ 2 out of 21 considering the skin as an additional class.

Abstract
This paper proposes perceptual segmentation of natural color images using a fuzzy-based hierarchical algorithm and its application to the segmentation of dermoscopy images. A fuzzy-based homogeneity measure makes a fusion of the color features and the texture features. The proposed hierarchical segmentation method is performed in four stages: simple splitting, local merging, global merging and boundary refinement. The effectiveness of the proposed method is confirmed through computer simulations that demonstrate the applicability of the proposed method to the segmentation of natural color imagesand dermoscopy images. ©2008 IEEE.

Abstract
A method to produce rectified Advanced Very High Resolution Radiometer (AVHRR) datasets is proposed, using both the orbital model and identification of ground control points (GCPs), involving a single image transformation from the raw imagery. The ability of the orbital model to determine the geographical co-ordinates of pixels in a raw AVHRR image was tested, using a total of 1098 GCPs in 24 AVHRR images. Five of the 24 images were also rectified using the method proposed, using a newly identified set of GCPs. The differences between the geographical co-ordinates of the existing GCPs and those determined by the new method are calculated and discussed.