Reserach On Image Calibration Algorithm For Fish-eye Lens

Fish-eye lens have a large field of view with a short focal length, which are widelyused in the full range of vision systems and other fields, but the fish-eye image videocamera has a serious image morphing. In order to recover the actual scene with theseserious deformed projection information, we must convert them to perspective projectionimages that suit human's vision. Therefore the fish-eye image distortion correctiontechnology has become a hot topic now.Fish-eye image distortion correction mainly includes three aspects of contents:fish-eye image pre-processing, contour extraction and image distortion correction.Firstly, we need take fish-eye image pre-processing before image corrections,because the results of preprocessing related to the post-processing of contour extractiondirectly and the quality of image calibration. Besides it can reduce the impact of noise,eliminate useless information, enhances and highlights the useful image information.Generally pre-processing is divided into color image binarization, image de-noising,morphological processing operations.Secondly, the current fish-eye image contour extraction method can not extract thecontour of images accurately, especially when there present large number of black pixelson the effective regional edge. This article poses an improved fish-eye image contourextraction method, compared with previous algorith, the proposed algorithm has moreextensively applicable scope and higher precision. Because there is a white aperture of theoutline edge on the picture that photographed in our laboratory, so the previous of variousalgorithms can not get the image contours accurately. Based on Hough circle detection,this paper proposed an algorithm which can accurate positioning fish-eye images.Thirdly, based on studying and analyzing variety of domestic and foreign fish-eyeimage correction method in the aspect of image distortion correction, this papersummarized two simple and rapid method: spherical coordinates positioning method andlatitude and longitude mapping method, so the fish-eye image distortion correction can be achieved. After correction of image morphing, the paper corrected image by interpolationcompensation which make calibration effect even better.Finally, because some observers are interested in the region that far from the opticalcenter which would resulted to a large distortion in the photograph, so fish-eye image ofthe optical center of the transfer algorithm have been studied, which can expend a fish-eyeimage into a series of small pictures.