Spectral Similarity Measure Edge Detection Algorithm in Hyperspectral Image

Hyperspectral remote sensing is a new and fast growing remote sensing technology that currently being widely investigated by researchers and scientists. Much of hyperspectral image analysis is focused on information extraction within a single pixel. However, information about the geometrical shape can improve the capability of recognizing ground truth as different kinds of targets with similar spectral. This paper focused on edge detection in hyperspectral image. Spectral similarity measures were introduced for the spectral feature variations in neighborhood and spectral similarity measure edge detectors were proposed. Then a spectral similarity edge detection algorithm was developed to extract edge information from hyperspectral image, which extends the traditional edge detection technique to high dimension of hyperspectral image. In experiments, spectral similarity edge detection algorithm demonstrated excellent performance in a real hyperspectral image.