Improved method of automatic image stitching based on SURF

Abstract

Image stitching is a process of assembling images of same scene into a large image. Traditional approaches usually have restrictions on the image sequence and stitching precision is low. For a better stitching result, especially in medical image processing, which needs more reliable and higher stitching precision, this paper proposes an improved method of automatic image stitching based on SURF (Speeded Up Robust Feature). In a mixed image set, we first use phase correlation to estimate if two images overlapped, and ascertain the relationship of the overlapping images, then adopt SURF algorithm to extract and match features in the overlapping parts, and at last improve image fusion strategy and use frame-by-frame expanded mosaic method to get panorama image. The experimental results show that our method is robust, which computes features with higher precision and can realize the smooth transition in images of same scenes.