In this paper, we propose a new method to estimate an infinite
homography between two views containing two arbitrary planar
rectangles. The proposed method does not require metric measurements,
such as rectangle lengths or aspect ratios of the rectangles. We introduce
the concept of semi-metric cameras and show that the semi-metric cameras
derived from different views that see an identical 3D rectangle, can
be regarded purely translating cameras whose pixel is zero-skewed. New
parameterization for infinite homography is developed based upon the
semi-metric space, and this parameterization is used to propose a new
algorithm to estimate infinite homography. As a direct application, we
apply our algorithm to autocalibration for a scene only with a few feature
points on each rectangles.
homography between two views containing two arbitrary planar
rectangles. The proposed method does not require metric measurements,
such as rectangle lengths or aspect ratios of the rectangles. We introduce
the concept of semi-metric cameras and show that the semi-metric cameras
derived from different views that see an identical 3D rectangle, can
be regarded purely translating cameras whose pixel is zero-skewed. New
parameterization for infinite homography is developed based upon the
semi-metric space, and this parameterization is used to propose a new
algorithm to estimate infinite homography. As a direct application, we
apply our algorithm to autocalibration for a scene only with a few feature
points on each rectangles.