Large-field 3D imaging using an array of gradually-changed FOV cameras

Large-field image stitching and 3D reconstruction are two hot topics in computer vision. Generally, only a small amount of overlapped area is required in large-field image stitching. On the other hand, cameras usually share almost the same field of view for multi-view 3D reconstruction, which guarantees enough overlapped areas to calculate 3D information of the target scene. To realize the two purposes using a single imaging device simultaneously, we propose a novel structure design of camera array, which combines multiple cameras with gradually-changed field of view. Based on the proposed camera array, we can firstly calculate the depth of points in 3D space, then a panoramic image with large field of view can be generated in consideration of the depth information. A prototype imaging system based on the novel camera-array structure is realized, and then real-scene imaging experiments are implemented to prove the feasibility and effectiveness of our method.