New HyperSpectral Image Segmentation based on the Concept of Binary Partition Tree

In this paper we have implemented a HyperSpectral Image Segmentation which is based on a new Binary Partition Tree pruning strategy which aimed at the hyper spectral images segmentation based on the concentrated object depth wise and effectively utilizing the sparse sources of the root cluster. The Binary partition tree is a region-based representation of images that involves a reduced the count of elementary primitive and therefore allows us to define robust and efficient segmentation algorithm. Recursive spectral graph partitioning helps to study the regions contained in the Binary Partition Tree branches. The aim is to remove sub trees composed of nodes which are considered to be similar and consider the entire sparse object. To this end, affinity matrices on the tree branches are calculated using a new distance-based measure. It has led to the use of such images in a growing number of applications, such as remote sensing, food safety, medical research etc. Hence, in the field of hyper spectral image segmentation, a great deal of research is invested. The number of wavelengths per spectrum and pixel per image as well as the difficulty of handling spatial and spectral correlation explain that why this approach is still a largely open research issue. The work proposed here focuses on the problem of image segmentation and the results obtained are efficient.