International audienceWe address the problem of unsupervised band reduction in hyperspectral remote sensing imagery. We propose the use of an information theoretic criterion to automatically separate the sensor's spectral range into disjoint subbands without ground truth knowledge. Our approach, named BandClust, preserves the physical sense of the spectral data and automatically provides relevant spectral subbands, i.e., of maximal informational complementarity. Experiments using real hyperspectral images are conducted to compare BandClust with four other unsupervised approaches. The comparison of the selected dimensionality reduction methods is performed via supervised classification using support vector machines and shows the potential of the proposed approach
International audienceWe present a new method for the visualization of spectral images, based on a selection of three relevant spectral channels to build a Red-Green-Blue composite. Band selection is achieved by means of information measures at the first, second and third orders. Irrelevant channels are preliminarily removed by means of a center-surround entropy comparison. A visualization-oriented spectrum segmentation based on the use of color matching functions allows for computational ease and adjustment of the natural rendering. Results from the proposed method are presented and objectively compared to four other dimensionality reduction techniques in terms of naturalness and informative content
International audienceWe introduce a new feature extraction model for purposes of image comparison, visualization and interpretation. We define the notion of spectral saliency, as the extent to which a certain group of pixels stands out in an image and in terms of reflectance, rather than in terms of colorimetric attributes as it is the case in traditional saliency studies. The model takes as an input a multi- or hyper-spectral image with any dimensionality, any range of wavelengths, and it uses a series of dedicated feature extractions to output a single saliency map. We also present a local analysis of the image spectrum allowing to produce such maps in color, thus depicting not only which objects are salients, but also in which range of wavelengths. A variety of applications can be derived from the resulting maps, particularly under the scope of visualization, such as the saliency-driven evaluation of dimensionality reduction techniques. Results show that spectral saliency provides valuable information, which do not correlate neither with visual saliency, second-order statistics nor with naturalness, but serve however well for visualization-related applications
We investigated nearest-neighbor density-based clustering for hyperspectral image analysis. Four existing techniques were considered that rely on a K-nearest neighbor (KNN) graph to estimate local density and to propagate labels through algorithm-specific labeling decisions. We first improved two of these techniques, a KNN variant of the density peaks clustering method dpc, and a weighted-mode variant of knnclust, so the four methods use the same input KNN graph and only differ by their labeling rules. We propose two regularization schemes for hyperspectral image analysis: (i) a graph regularization based on mutual nearest neighbors (MNN) prior to clustering to improve cluster discovery in high dimensions; (ii) a spatial regularization to account for correlation between neighboring pixels. We demonstrate the relevance of the proposed methods on synthetic data and hyperspectral images, and show they achieve superior overall performances in most cases, outperforming the state-of-the-art methods by up to 20% in kappa index on real hyperspectral images.
Even though the study of saliency for color images has been thoroughly investigated in the past, very little attention has been given to datasets that cannot be displayed on traditional computer screens such as spectral images. Nevertheless, more than a means to predict human gaze, the study of saliency primarily allows for measuring informative content. Thus, we propose a novel approach for the computation of saliency maps for spectral images. Based on the Itti model, it involves the extraction of both spatial and spectral features, suitable for high dimensionality images. As an application, we present a comparison framework to evaluate how dimensionality reduction techniques convey information from the initial image. Results on two datasets prove the efficiency and the relevance of the proposed approach.
We propose a new strategy to evaluate the quality of multi and hyperspectral images, from the perspective of human perception. We define the spectral image difference as the overall perceived difference between two spectral images under a set of specified viewing conditions (illuminants). First, we analyze the stability of seven image-difference features across illuminants, by means of an information-theoretic strategy. We demonstrate, in particular, that in the case of common spectral distortions (spectral gamut mapping, spectral compression, spectral reconstruction), chromatic features vary much more than achromatic ones despite considering chromatic adaptation. Then, we propose two computationally efficient spectral image difference metrics and compare them to the results of a subjective visual experiment. A significant improvement is shown over existing metrics such as the widely used root-mean square error.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.