摘要：This paper describes a novel method of real-time virtual view generation for augmented virtuality system. The aim is to rectify the displacement between virtual eye position and video cameras which are attached on the head mounted display. It consists of three steps: Firstly, stereo calibration was used to generate cameras' internal and external parameters, and then image planes were rectified to canonical configuration state that made them alignment or corresponding points on the same scan lines. At last the virtual view images between two cameras are generated using view morphing which is a shape-preserving transition method from a source to a destination view. We can adopt this method to choose locations of two virtual cameras yielding an in-between stereoscopic camera views without reconstructing the 3D model of the object. The view of the object corresponding to a virtual camera that is located between two real cameras with same position of eyes. So we can control the feeling of depth and provide customized viewpoint to the user. It can be used in a variety of augmented virtuality applications to integrate live video objects into virtual environments. Experimental results verify the validity of the proposed approach, and we also show interlaced stereo composition results using a pair of synthesized views from input views acquired by stereo cameras.
