Video Mosaicing for Document Imaging

This paper describes a real-time video mosaicing method for digitizing paper documents with mobile and low-resolution video cameras. Our proposed method is based on the structure-from-motion technique with real-time processing. Therefore, it can dewarp an image suffering from perspective deformation caused by a slanted and curved document surface. The curved surfaces are reconstructed as an expanded flat image by 3D polynomial surface regression and projection onto a flat mosaic plane. The order of the polynomial is automatically determined according to the geometric AIC. Moreover, the real-time scanning and reconstruction process offers interactive user instruction on the style of user’s camera handling, such as appropriateness of camera posture and scanning speed. Manual camera scanning for video mosaicing captures a point on the target multiple times and that allows other merits. They include super-resolution and shading correction. The experiments demonstrated the feasibility of our approach.