Research And Application On Technology Of Image Mosaic

Image mosaic is a technology which can combine two or more small-scale images which come from the same scene and have overlap region into one image with a large view. Image mosaic is an important branch of image processing which can overcome the limitations of shooting angle and has been widely used in remote sensing image processing, medical image analysis, cartography, computer vision, video surveillance, virtual reality and super-resolution reconstruction and so on. In this paper, some studies are done in two key technologies of the image mosaic: image registration and image fusion, and corresponding algorithm is put forward. At last, we have a deep analysis with the problem of multiple image mosaic.Compared to the classic corner detection algorithm, a new algorithm based on delaminateion matching of corner-points for multi-image mosaic matching is presented in this paper. The algorithm takes the idea of matching step by step. First, we use Harris Corner Detector to extract feature points. Second, rough matching points are obtained through normalized cross-correlation and the false matching points are filtered according to the edge information around the feature points. Third, we refine the matching points using RANSAC. At last, we get a complete image with high quality using the method of minimum weighted distance for fusion. Experimental results show that this algorithm can accurately find overlapped region between images, and the speed and applicability of mosaic are comparatively satisfied. But the results will be affected by the content of image itself and the stability of this algorithm need to be improved.To further enhance the stability of the algorithm so that it can meet the demands for more mosaic images, a new algorithm of image mosaic based on scale-invariant features transformation(SIFT) is proposed in this paper. Firstly, we uses the algorithm of SIFT to detect feature points, and then combines the cosine similarity and Euclidean distance to filter and match the feature points several times to get the matched-points with higher accuracy, and finally introduces the method of minimum weighted distance for fusion. This method of feature matching based on SIFT can get more rich information for each point, and the matching way of feature points is more flexible. Compared with general method, it has smaller sensitivity to noise and illumination, and higher stability.The image mosaic based on scale-invariant features transformation(SIFT) is applied in fabric image. This algorithm consider the first entered image as the standard coordinate space, and other images are transformed to the standard space according to the transformation between adjacent images. In order to achieve precise fabric image mosaic, we have achieved highly accurate matching points through the eigenvector choice according to the characteristics of fabric image itself. Experimental and analytical results show that the results of image mosaic receive a high subjective evaluation and also can meet the most of human's visual requirements well.Finally, we overview the development of the image mosaic and the research of this paper, and also discuss the issues which have not resolved but have worth for study.