Hemi-cylinder unwrapping algorithm of fish-eye image based on equidistant projection model

This paper presents a novel fish-eye image unwrapping algorithm based on equidistant projection mode. We discuss a framework to estimate the image distortion center using vanishing points extraction. Then we propose a fish-eye image unwrapping algorithm using hemi-cylinder projection in 3D space. Experimental results show that our algorithm is efficient and effective. In particular, the hemi-cylinder unwrapping results do not reduce the horizontal field of view which is very useful for panoramic surveillance with applications in important sites safety compared with other fish-eye image correction methods.