A Stereo Vision through Creating a Virtual Image Using Affine Transformation

We propose a new approach for stereo vision. In human vision one percept can arise from two retinal images as a result of the process called "fusion". We try to create a virtual image from two images of stereo cameras. Optical axes of the cameras intersect on an attention point locating on the surface of an object. By Affine transformation of the respective images, the two images are transformed to a virtual image that is seen from the middle point between two cameras. By superimposing two transformed pictures, one fusion picture of virtual image is reduced. Excepting for around the attention point, two transformed pictures do not coincide with each other. Finding correspondence between two pictures, we can determine Sdimentional depth information.