Camera Rotation Invariance of Image Characteristics

摘要

The image transformation due to camera rotation relative to a stationary scene is analyzed, and the associated transformation rules of 'features' given by weighted averaging of the image are derived by considering infinitesimal generators on the basis of group representation theory. Three dimensional vectors and tensors are reduced to two dimensional invariants on the image plane from the viewpoint of projective geometry. Three dimensional invariants and camera rotation reconstruction are also discussed. The result is applied to the shape recognition problem when camera rotation is involved.