Research Of Image Mosaic Technology Based On SIFT Feature

Image mosaic technology is one of the hot spots in the field of computer and image processing,the object is to splice the images with overlapping parts in the same scene,to get a picture that contains all the information of the image, with high resolution and wide field of view. Image mosaic technique overcomes the contradiction between wide field of vision and resolution. It is widely used in the field of the panoramic image synthesis, the scene of the 3-D reconstruction, medical image analysis, video surveillance system, aviation, satellite image integration and so on. This paper is based on the scale invariant feature transform (feature transform Scale-invariant, SIFT) to achieve image mosaic technology. Image stitching technology includes image acquisition, image preprocessing, feature extraction, image registration and image fusion of the several parts. The main work done in this paper includes:(1) Research on the theory of image preprocessing. The image preprocessing algorithm commonly used are studied, analyzed and compared to the commonly used image preprocessing algorithms, in this paper, the median filter is used to remove noise, and the histogram equalization method is used to enhance the image. The basic theory of image registration and image fusion technology is studied.(2) Research on the algorithm of extracting common feature points. Comparing the advantages and disadvantages of extracting common feature points algorithm, this thesis selected SIFT feature extraction algorithm as the basis. After extracting feature points, the Kd-tree based BBF algorithm is used to match the feature points, and then the RANSAC algorithm is used to eliminate the error matching. Aiming at the problem that the RANSAC algorithm has a large number of feature points, the time of the operation of the algorithm will be increased dramatically due to the random selection of initial matching pairs. The RANSAC algorithm is improved when the initial sample points are selected. Experimental results show that the improved algorithm can improve the efficiency of operation and save time, which can meet the real-time performance of the algorithm.(3) Image fusion processing. The image fusion technique is used to deal with the final result, and the experiment results show that the algorithm can effectively remove the shadow in the overlapping area of the mosaic image. The reliability of the proposed algorithm and the performance of the improved algorithm are verified by experiments with three different sets of conditions.(4) In order to reflect the specific process of image mosaic, an image mosaic software is developed.