Hyperspectral Anomaly Detection via Dual Dictionaries Construction Guided by Two-Stage Complementary Decision

Low rank and sparse representation (LRSR) with dual-dictionaries-based methods for detecting anomalies in hyperspectral images (HSIs) are proven to be effective. However, the potential anomaly dictionary is vulnerable being contaminated by background pixels above methods, this limits effect of detection (HAD). In paper, a dual dictionaries construction method via two-stage complementary decision (DDC–TSCD) HAD proposed. first stage, an adaptive inner window–based saliency was proposed yield coarse binary map, acting as indicator select pure pixels. For second estimation network designed generate fine map. Finally, map worked together construct guidance superpixels derived from stage. The experiments conducted on public datasets (i.e., HYDICE, Pavia, Los Angeles, San Diego-I, Diego-II Texas Coast) demonstrate that DDC–TSCD achieves satisfactory AUC values, which separately 0.9991, 0.9951, 0.9968, 0.9923, 0.9986 0.9969, compared four typical three state-of-the-art methods.