Research On Image Matching Technique

Image matching is an important research topic in the fields of computer vision and image processing. It used to match two or more images, for example, at different times, from different sensors or from different view points. Image registration is the foundation of many image processing and applications. The effect of registration directly works on the following processing steps.A great deal of work is done to study image matching algorithms in this paper. Based on the analysis of the edge detection and feature matching algorithms, an improved edge detection algorithm has been summed up. This algorithm can automatically select the scale.According to the characteristic that the edge feature points are relatively constant to the spatial location, a new image matching algorithm which based on edge feature points has been proposed. The method firstly extracts the edge feature points by using of polar coordinates. And then it creates the direction-angle chain codes. Secondly, it uses the information of chain codes to find the matching edges. Finally, the shift and rotation has been calculated. The experiment result shows that this method can deal with the images existing shift and rotation well.Because of the limitation of the proposed algorithm, an improvement algorithm has been implemented by combining the algorithm of interpolation and B-Spline theory without changing up the original performance. The advanced matching algorithm has greater precision and also applied to the images existing scale. The method extends the scope of application and has the excellent matching results.Lastly, the lack of the research in the paper is summarized, and then looks forward to the futurity of the research of image matching.