Automated clustering/segmentation of hyperspectral images based on histogram thresholding

Segmentation of multi-dimensional infrared imagery from histograms

“…We observed that increasing this value would have striking effects on the consequent segmentation (see for example Fig. 4(c) and (d) in [9]). However, choosing the peak definition that gives the most pleasing segmentation is both subjective and empirical and we prefer the alternative scaling technique outlined next.…”

“…2 and 3 incorporate some recent refinements to the simple technique described above [9] which address two issues. First, the nearest peak as measured by Euclidean distance used in forming the template (Fig.…”

“…In our initial formulation of the histogram segmentation technique [9], we used linear scaling and iterated the specified range of integer bins until the user-desired number of segmentation levels was achieved. The only other user choice was to fix the minimum number of pixels at a maximum histogram location accepted as a peak (default ''2'' used initially).…”

Evaluating Subpixel Target Detection Algorithms in Hyperspectral Imagery

“…For the ACE detector, the same procedure (replacing m g by m 8 ) may be followed. Segmentation [16][17][18] or even more local covariance matrices [2,4,6,19] can be used to improve the covariance matrix. Common to all these methods is an increased need for high performance computational resources, while the corresponding influence each method has on detection ability is uncertain and highly dependent on the pictures being analyzed.…”

A User-Assisted Approach to Visualizing Multidimensional Images

“…The pixel values in a multiparameter image can be quantized into a displayable number of colors using unsupervised segmentation algorithms [16], [17]. The main drawback of these approaches is that they may fail to preserve intracluster variation and the process of assigning colors to clusters either is a difficult manual process or relies on heuristics.…”