Research Of Geometric Correction And Stitching Algorithm Of Fisheye Image

Fisheye lens a kind of super wide lens, and can meet or exceed the perspective of the human eye. General perspective of the fish-eye lens close to or equal to 180? the most can reach 220皁r230? However, due to the extremely short fisheye lens focal length, the distortion caused by optical principle is the more serious. This distortion is not suitable for human eye’s observation, and brings some obstacles for fisheye lens’ application, therefore, the correction of distortion is very important. This paper has done a detailed study about many methods. First, introduce simply the imaging principle and acquisition method of fisheye lens, classify the fisheye images based on different perspectives and different shooting position of lens, and analyze the causes of image distortion and types of distortion.According to classification of fisheye images, study the different models of fisheye image correction. For the correction of fisheye image that vertical viewing angle is 180(?), introduce algorithms based on the cylinder model and the spherical model, then propose an improved algorithm based on the spherical model, and give the detailed steps of the method. The experimental results show that this new algorithm is satisfactory, and has the faster computing speed. About the panorama spread using one fisheye image that horizontal viewing angle is 360, introduce the expansion algorithms based on half-cube model and cylinder model, and propose a new method of panorama spread with one 360? fish eye image. And compare the experimental results between the new algorithm and the algorithm based on Half-cube model and cylinder model. The result of comparison show that the improved algorithm is not only faster, but also realizes the spread of all pixels in image and makes up for shortcomings of the half cube model and cylinder model, which have the blind area in the expand panorama. According to angular fisheye projection and mirror principle, this paper presents a new correction algorithm with fisheye images, whose vertical viewing angle is more than 180? This algorithm can correct the part of more than 180? in fisheye image effectively, and makes up the vacancies in this area. Sometimes, do not need to observe the entire scene, only need to observe local. So this paper presents local correction algorithm based on sundial projection. And make it to a software system.In second part of this paper, introduce image mosaic with corrected fisheye image. Analyze advantages and disadvantages of various image matching algorithms, then chose SIFT algorithm to match different images. Describe in detail the principles of SIFT algorithm, and experimental results show that this algorithm can be more accurate and less error in feature points match. Finally, implement the image stitching on the basis of SIFT algorithm.