Stitching Of Panoramic Images Based On SURF

The panoramic image has been widely used in virtual reality,computer vision, aerospace,military applications,medical image analysis,target identification and tracking,remote sensing image processing and other fields,with a wide viewing angle and high authenticity. Image matching is the important in panoramic image stiching. So,This paper focuses on the image matching technology based on feature points, try to find a feature extraction algorithm with faster and higher matching accuracy.The Image matching algorithm based on feature points mainly includes two main types of image matching, based on corner feature and based on the invariant feature points matching. First we introduced the image matching algorithm based on corner features, Harris algorithm, for example, Experimental analysis carried out to verify the performance of this kind of matching algorithm. The algorithm is not stability, so that is suitable for the picture of the obvious features of angular point. Then study on the invariant feature point matching algorithm. Focus on performance testing and analysis of the characteristics of the SIFT algorithm of classic match. SIFT algorithm has good robustness and anti-interference, real-time performance is not high. The SURF algorithm proposed on the basis of the SIFT algorithm, has a good real-time performance and reliability is not high. Finally, SURF algorithm based on reliability test is proposed to improve the reliability of the algorithm. In order to improve the matching speed, fast matching algorithm is proposed, using vector dot product to replace the Euclidean distance, greatly reduces the amount of calculation. The experimental results show that the improved SURF algorithm and fast matching algorithm hold good robustness and high matching speed. The algorithm has a high practical value especially in scene of higher real-time demand.Finally, in the realization of fast and accurate registration on, and further study of image fusion and cylindrical projection transformation, and ultimately by improved algorithm stitching cylindrical panoramic images.