Typhoon cloud image retrieval is very important for typhoon research and forecast. In this paper, we designed a multifeature
typhoon cloud retrieval system. Using this system, we investigated eight Content-based Image Retrieval (CBIR)
methods, including histogram-based, texture-based, shape-based and entropy-based retrieval methods. The performance
of these methods was tested by three experiments from different aspects. They are paradigm retrieval experiment,
random statistical validity experiment and random statistical efficiency experiment. The experimental results indicate
that the validity and effectiveness of entropy-based retrieval method is the best in all the methods. And some methods
(like HU moments-based image retrieval method) are not suitable to the typhoon cloud images retrieval.