Research And Implementation Of 3D Panoramic Image Correction And Feature Extraction Algorithm

Panorama, a virtual 3D scene, is achieved by processing multiple fisheye images and we are able to introduce lots interactive actions on this basis. It gives visitors a feeling of true scene. Compared with 3D modeling, three-dimensional panoramic system has a faster speed and more real in the construction of the scene. Because of these characteristics, three-dimensional panorama technology has been applied in various fields, such as panorama maps, browse of attractions and virtual house. As more and more applied in the field of three-dimensional panorama technology, the quantity of fish eye image is increasing dramatically. The traditional algorithm of fisheye image correction and image mosaic hasn’t suitable for today’s fast developing society. Therefore, improve the efficiency and flexibility of the algorithm is imperative.Fisheye image correction and image mosaic are two important problems need to be solved in three-dimensional panorama technology. On this paper, we study the basic principle of the imaging process of fisheye image and algorithms of fisheye image correction. On the basis of the original algorithm, an improved program is put forward, which is efficient and controllable. The algorithm is able to adapt to the different degrees of the fisheye image distortion by introducing the distortion correction parameter. On the other hand, this paper introduces a new key point descriptor by studying the basic principle and implementation details of SIFT and the new key point descriptor which effectively reduces the dimension from 128 to 96.Finally, through the experiment, we confirmed the improvement of the fisheye correction algorithm has good controllability, and the new feature point descriptor improves the efficiency while guarantee the stability of the algorithm effectively.