Research And Implementation Of SAR Image Retrieval System Based On The Nonsubsampled Contourlet Transform

With the development of Synthetic Aperture Radar(SAR) technology,the resolution of SAR image is equal to optics image,and has richer information. The application of SAR image has become more and more widely. How to retrieve the right one among numerous images efficiently and quickly becomes an important branch in the field of image's application. In order to manage and retrieve large amounts of images, the Content-based image retrieval (CBIR) has emerged to be one of the hot research areas in image domain.In this thesis, we mainly research on the application of the multiscale geometric analysis the Nonsubsampled Contourlet transform in image retrieval algorithms based on texture feature and shape feature. Based on the system of radar image database design, we implement the SAR image retrieval system on Oracle database. The main content of this thesis are summarized as follows:1. First we analyze and study the principle of image retrieval system and key techniques and algorithms of CBIR, such as the low-level feature descriptions including texture, shape, the similarity measure between images, the indexing methods and so on2. Researching on the texture-based image retrieval algorithm, we propose an algorithm of texture feature extraction based on the Nonsubsampled Contourlet transform in this thesis. The image is decomposed by the Nonsubsampled Contourlet transform. The mean, standard deviation and third central moment of the magnitude of the Nonsubsampled Contourlet coefficients at different scales and directions are computed to extract the texture feature vector.Experiment proves the third central moment added in NSCT arithmetic is overperformded than only use the mean and standard deviation, and precision ratio has improved.3. Researching on the shape-based image retrieval algorithm, we propose an algorithm of shape feature extraction based on the Nonsubsampled Contourlet transform in this thesis. Combining the improved canny oper- ator with the NSCT, we extract the edge of the image with the improved canny operator; then use NSCT decomposing the shape information to different scales and different directions. As a result, shape information under each frequency is remained and loss is reduced.4. We analyze the working of spaceborne SAR groud processing system, clarify the importance of groud database management system in the whole SAR ground processing system.Design the pigeonhole, searches, browse of radar data, and complete the design of ueser's database tables.5. Finnally we design a primary content-based SAR image retrieval system using Visual C++ 6.0, Matlab7.0.4 and Oracle 9i. A SAR image retrieval system is implemented by using the texture and shape extraction algorithms based on the Nonsubsampled Contourlet transform. The system can not only implement the texture and shape retrieval of SAR images, but also be used for algorithm evaluation and performance comparison. Besides, different algorithms can be combined to get better results.