Reclustering hyperspectral data using variance-based criteria Conference Paper uri icon

abstract

  • Abstract: We have examined the clustering results obtained via our previously published N- dimensional histogram segmentation algorithm. In particular, we have derived a method to recombine areas that have been oversegmented in the initial segmentation process. While the algorithm does reduce the number of clusters, different initial clustering inputs do lead to different clustering results. Methods to compare the different final segmentations will be discussed.

publication date

  • January 1, 2004