Image Stitching Algorithm Based On SURF

Image Stitching is a very important application in image processing, in recent years, image stitching technology has a very practical application in remote sensing imaging, computer vision and medical imaging. Image stitching process has the following five components:(1) pre-processing the image; (2) image registration; (3) to determine the splice transformation model; (4) to be spliced image conversion; (5) image fusion.Image registration plays an important role in the whole splicing process, and directly impacts the stitching result. The problems of low accuracy rate and slow registration speed exist in existing registration algorithms. In order to improve the accuracy of image registration and stitching speed, this paper focuses on classic feature point detection algorithm and matching algorithm based on feature points, and gives some improvements on the basis. Existing improved image registration algorithms improved feature detection algorithm or precise matching RANSAC algorithm used in the process in generally so as to improve the registration rate effects. This article used the existing feature point features to achieve multi-feature fusion matching, increased registration accuracy, relative to other improved algorithm has better improved results.This paper introduces the image stitching algorithm commonly used, and improve the original FLANN registration algorithm based on the classical SURF algorithm, some existing features of SURF are applied to registration screening. It gets the realization of multi-feature fusion match and improves the accuracy of registration.Next, FAST algorithm has a significant advantage in feature point detection speed,this paper uses FAST algorithm to the image stitching, and contrasts the stitching speed of FAST and SURF algorithm. Experiments show that, the stitching speed based on FAST is faster 2-3 times than the stitching speed based on SURF. In addition, FAST algorithm has stable performance and can achieve a good image mosaic effect.