A NOVEL HYPERSPECTRAL ANOMALY DETECTOR BASED ON LOW-RANK REPRESENTATION AND LEARNED DICTIONARY

The low-rank property of hyperspectral imagery is well exploited with low-rank decomposition methods recently. In our approach, a novel hyperspectral anomaly detector based on low-rank representation (LRR) and learned dictionary (LD) has been proposed. This method assumes that a two-dimensional matrix transformed from a three-dimensional hyperspectral imagery can be decomposed into two parts: a low rank matrix representing the background and a sparse matrix standing for the anomalies. Further, a dictionary learned from the whole image with a random selection process, which can be viewed as the spectra of the background only, is introduced into the decomposition process. The adoption of LD improves the robustness of LRR to its parameters with a less computational cost. Experimental results demonstrate that the proposed method has a satisfactory result.