A Deformation-minimized Reprojection Method Based on Generalized Planar Rectification

In order to be in favor for image registration and mosaic in computer vision, a planar rectification on epipolar geometry is studied, which minimizes the deformation including the projection distortion and the resampling effect. Firstly, a pair of images is normalized to simplify fundamental matrix, and the geometric constraint is applied to simplify rectification transform matrix. Secondly, the extremum of sum of projective transform weight change is solved to minimize projective distortion. Lastly, the rectification transform matrix is solved to minimize resampling effect that is weighed by Jacobian determinant. And a deformation-minimized reprojection idea based generalized planar rectification is proposed to deal with resampling effect from rotation. The generalized planar rectification coordinate system is built and normalized hypothetically. Then, minimizing distortion and resampling effect is calculated in the hypothetical normalized coordinate system. It keeps the similarity before and after transforming, and the visibility is improved. It is in favor of the succedent image registration and mosaic