This thesis thoroughly studied the technique of Region-of-Interest (ROI) Based Image Retrieval, expatiated the new method that based on human-eye vision characteristic to acquire ROI in detail, and improved the technique of Region-of-Interest (ROI) Based Image Retrieval.For the ROI acquisition, this thesis use the subregion based region grow method to increase the speed and anti-jamming ability. Besides the saliency map and relative position, this thesis add the image content (color and texture) to acquire the interested region .This method can acquire the ROI more integrated.For the image character describe, this thesis not only describe the ROI but also describe the holistic character .It can weaken the infection of inaccurate ROI acquisition, enhance the retrieve accurate.By tool of Visual C++ and Matlab simulate the result of research, as a result, this thesis's arithmetic can gain better effect.
|