Mixture gradient detector for subpixel detection

Subpixel detection is an important but difficult problem in hyperspectral image. Due to the small size of the target, only spectral information can be used for detection. Many algorithms have been proposed to reduce this problem, and most of them assume that the distribution of hyperspectral image is multinormal. However, this assumption may not be an appropriate description of the distribution in hyperspectral image. After carefully study the distribution of hyperspectral image, it is concluded that the gradient of noise should also be considered. In this paper a new model is proposed, which assumes that gradient of the noise also follow Gaussian distribution. Based on the given model, two detectors, mixture gradient structured detector (MGSD) and mixture gradient unstructured detector (MGUD) are proposed. The proposed detectors take advantage of the new model, in which the distribution of noise is more accordant with the practical situation. Experiment results demonstrate that in general the proposed detectors perform better than state-of-the-art.