Panoramic Image Mosaic Based On SURF Feature Matchning Algorithm

With the increase of application requirements, image mosaic technology has become an important research subject on the field of virtual reality, computer vision, computer graphics, video processing and so on.This paper mainly studies the image mosaic technology based on feature points. Seeking a highly effective registration method based on points feature is the research core. Firstly, the image matching based on corner feature is studied systematically. Registration performance analysis and verification of the algorithm were also carried on. Furthermore, a systematic study on feature detection based on invariant methods was conducted. A detail analysis on the most popular SURF feature detection algorithm is discussed referring to the classic SIFT algorithm. On this basis, this paper designs and implements a image matching technique based on invariant methods, and makes performance verification of the SURF algorithm through experimental study. The results show that algorithm based on SURF not only has good robustness, well noise immunity and statistical characteristics, but also has a four times faster matching speed than SIFT.On the basis of the previous work, cylindrical panoramic image mosaic based on SURF was designed and realized finally.