SAR Image Retrieval Based On Gaussian Mixture Model Classification

With the development of computer technology, the format of information has changed into picture, graph and video from the single text. Moreover, the volume of that increased explosively. Enormous costs involved in the manual analysis of these large volumes have made it necessary to develop automatic image analysis and mining tools. The content-based image retrieval comes out consequently. It breaks the limitation of traditional retrieval methods and compares the similar of two images by the contents direct. Content-based image retrieval is the innovation in data mining.SAR image retrieval, lacking of well retrieval results recently due to the particularity, such as noise, of SAR image, has drawn more and more attention with the surprisingly increasing volume of SAR data and the dramatically enlarging application range of SAR image during almost half of the last century. This paper considers the characteristic of content-based image retrieval (CBIR) and SAR image together, proposing a method of SAR image retrieval. The proposed method can divide into two parts, including image classification and matching. Firstly we use Gaussian Mixture Model (GMM) to gain a precise result of classification, and then get the retrieval results through some matching algorithms which are good for image retrieval. Experimental results show that the proposed method can retrieve SAR images containing all kinds of surface features effectively.