Research on Different Feature Matching Algorithms for Panoramic Image Stitching
Zhao Zhang · Advances in computer science research · 2024
Panoramic image stitching technology has penetrated into every field of modern life.As an important part of the stitching process, image feature matching directly affects the quality and speed of the stitching.In this paper, photos taken in daily life are used for experiments, and the precision and computational efficiency of three different feature matching algorithms, Scale Invariant Feature Transform (SIFT), Speeded-Up Robust Features (SURF), Oriented Fast and Rotated BRIEF (ORB), under rotation, scaling, light intensity transformation and perspective transformation, are compared to explore their applicable scenarios.The experimental results show that SIFT is most appropriate for perspective transformation, but its running speed is so slow that it is only suitable for occasions where the real-time requirement is not high.SURF has the greatest stability when dealing with scale changes and different light intensities, while it operates far quicker than SIFT.ORB exhibits the best robustness in the case of rotation and runs the fastest in all cases, so it is most suitable for applications in real-time scenarios.