In order to overcome the low efficiency of anomaly detection for hyperspectral images based on sparse representation, a joint spatial preprocessing and spectral clustering based collaborative sparsity anomaly detection algorithm is proposed, which makes full use of the spatial and spectral properties of hyperspectral images, and establishes a cooperative processing mechanism between the spatial and spectral properties, according to the imaging principle and structure of the hyperspectral imagery. The spatial properties of the hyperspectral images are analyzed, and the spatial preprocessing is combined with the spectral properties, which makes the anomalous targets in hyperspectral images easier to be detected. Then, the spectral clustering method based on spectrogram division is used to divide the band subsets, and the spectral clustering method has the features of convergence to the global optimal solution and fast speed. The anomalous targets in each band subset are detected with the proposed new space and spectral collaborative sparsity divergence index method. This collaborative sparsity method considers the spatial and spectral properties of the hyperspectral imagery. Final anomaly detection result is obtained by the superposition of the results of each band subset. The real AVIRIS and synthetic hyperspectral imagery data sets are used for simulations. Simulation results demonstrate that the proposed algorithm is robust, and has higher precision and lower false alarm probability.