Many civil and military applications require fast and accurate detection of abnormal objects in hyperspectral images. In this paper, we present a novel, clustering-based algorithm called ADHSI-SD - Anomaly Detection in Hyperspectral Images with Spatial DBSCAN, which partitions a hyperspectral image into segments utilizing the spectral characteristics of the image pixels and then evaluates the anomality level of the pixels belonging to each segment. Time series analysis is used to extract and select the most relevant features from the spectral curve of each pixel. Pixel clustering builds upon a novel, computationally efficient combination of spatial and spectral features. In evaluation experiments on 14 real-world hyperspectral images from several domains, the proposed algorithm demonstrates a better detection performance than the classical Reed–Xiaoli (RX) anomaly detectors while shortening the detection runtime significantly.