Non-linear method used for distortion correction of fish-eye lens: Comparative analysis of different mapping functions

A fish-eye camera equipped with fish-eye lens is capable of providing a wider field of view (close to 180°), which makes it useful for surveillance, security, flight simulators, visual combat simulators, hemispherical photography, astrophotography and ADAS (Advanced Driver's Assistance System) widely. Though such lenses are considered a key component of the next-generation of cameras, the frames captured suffer from high scale distortion. Thus, fish-eye lens distortion correction techniques and algorithms have been developed to remove such distortions, when required. This paper presents a generalized method for generating undistorted images taken by cameras having fish-eye lenses mounted on walls, ceilings, floors and vehicles at any pitch or tilted angle. Pitch here refers to the angle of the camera lens from the horizontal. The proposed approach comprises of mainly two steps, viz. correcting the fish-eye lens images to rectilinear images, and rescaling the generated output image to the same aspect ratio as of input image. The images captured by the used fish eye lenses possess barrel distortion, for correction of which a non-linear and non-iterative method is used.