Fast Correction and Stitching Algorithm for Fisheye Images

Abstract

Fisheye cameras, with their ultra-wide-angle field of view, can capture a larger scene in a single shot compared to traditional lenses. This capability makes them highly valuable in areas such as security surveillance, panoramic photography, autonomous driving, and robot navigation. However, building an efficient and accurate visual system is crucial for realizing these applications. This paper proposes an improved algorithm based on fisheye image correction and stitching. First, a fast fisheye image distortion correction method is introduced to eliminate the severe distortion caused by the fisheye lens. Then, the o-FAST algorithm is used for feature point detection, ensuring rotational invariance of the feature points, and the KNN algorithm is employed for initial matching. On this basis, the PROSAC algorithm is applied to remove incorrect matches and compute the homography matrix between images. Finally, seamless image stitching is achieved using a gradual weighted fusion method. The research results demonstrate that this method significantly improves both image quality and computational efficiency, making it effective for generating and processing panoramic images. Experimental results further confirm that the method enhances stitching accuracy while also increasing processing speed, showcasing its high practical value.