Researches On Robust Image Matching Technology

Image matching has become an important technique in information processing area, it has been widely applied in various fields,such as industrial products test,military fields,remote sensing,data fusion, medical diagnose and other applications in computer vision.Image matching is a process designed to make two images aligned in space, or locate the corresponding target in the image according to a known pattern,called template image. Due to the complexity of the imaging condition, it is difficult to achieve the perfect results in terms of time consumption and matching precision. Therefore, in this paper, we are intending to improve the robustness of the matching algorithms.To address the issue caused by repetitive patterns,a robust method base on progressive matching is proposed. First,a rough homography matrix is established through matching with high saliency features based on saliency analysis. In the subsequent matching process, the projection error caused by the homography matrix is utilized to narrow the size of search region gradually to enhance the saliency of feature in the corresponding region.Simultaneously, the matching region is reduced with the same method. Revise the homography matrix whenever the matching is done with salient feature in the local region.Experimental results show that the proposed method significantly improves both the number of matching score and the correct matches.To solve the rotation problem in traditional template matching algorithms,a fast rotation invariant template matching method based on two-stage search strategy is proposed. The algorithm consists of two stages. In the first stage,a ring projection method is applied to select matching candidates. The hierarchical search strategy is adopted to reduce the computational cost for this stage. In the second stage,rotation invariant template matching is performed only on the matching candidates using method based on orientation codes. The experiment results demonstrate that the algorithm significantly improves the matching performance, especially in terms of time consumption.