Automatic estimation of distortion coefficient for correcting radial distortion

Abstract

This study proposes the facial landmark points based distortion coefficients estimation to overcome the radial distortion. Existing methods estimate the distortion rate using the predefined information, such as design model of lens, pattern information, and prior learning. The extracted scene information based on the proposed method successively correct the distortion in distortion image without predefined information and significantly improves the predefined step of distortion correction method.