The Research And Implement Of Correction Algorithm For Ultra Wide-angle Distortion Image

Ultra wide-angle lens, also called fish-eye lens, enjoy the large angle of field as their significant advantage. Thus, they have been widely used in various fields such as panoramic shooting, group video meeting, video surveillance, intelligent traffic system, robotic navigation and so on. However, it will bring about serious distortion when images are captured by ultra wide-angle lens. The distortion doesn’t fit the visual habits of people. Therefore, when we further use those images with distortion, it is significant for those images to be corrected to perspective projection images. Consequently, it is worth of doing research of correcting ultra wide-angle distortion images.The thesis first describes the purpose and significance of the project, analyses and summarizes the research situation of ultra wide-angle distortion image correction. Then relevant theoretical knowledge of image correction is briefly discussed, such as imaging principle of traditional lens, classification of lens distortion and fish-eye lens distortion models. Next, some kind of extraction of fish-eye image circular effective area and experimental results are given with strengths and weaknesses from adaptability, accuracy and computational amount. Meanwhile, an improved algorithm of extraction is proposed with verification its accuracy and comparison with other algorithm. After that, some kinds of fish-eye image distortion algorithms are introduced, including polynomial correction algorithm based on eigenvalue minimization, correction algorithm based on longitude model and correction algorithm based on contents. Besides, the thesis gives the strengths and weaknesses of these kinds of algorithms and proposes an improved correction algorithm based on longitude model. Finally, because of some pixels need to be interpolated after correction, some kinds of image interpolation algorithms are introduced, including traditional nearest interpolation, bilinear interpolation and bicubic interpolation. In order to improve blurring in detail region, implicit image interpolation algorithms based on edge are introduced. Further more, an improved implicit image interpolation algorithm is proposed.