View morphing based on auto-calibration for generation of in-between views

초록

Since image morphing methods do not account for changes in viewpoint or object pose, it may be difficult even to express simple 3D transformations. Although previous view morphing can synthesize change in viewpoint, it requires camera viewpoints for automatic generation of in-between views and control points by user input for post-warping. This paper presents a new morphing algorithm that can generate automatically in-between scenes by using auto-calibration. Our method rectifies two images based on the fundamental matrix, and computes a morph that is linear interpolation with a bilinear disparity map, then generates in-between views. The proposed method has an advantage that knowledge of 3D shape and camera settings is unnecessary.