3D Integral Imaging Using Sparse Sensors With Unknown Positions

In this paper, the 3D integral imaging method is extended to situations where the sensor positions in the image pick up stage are unknown. Conventional integral imaging systems require a priori knowledge of sensor positions in the image capture stage which can be difficult to measure in synthetic aperture or randomly distributed sensors modes. In the proposed method, only the relative position of two sensors is needed whereas all other sensors positions are unknown. We combine image correspondences extraction, camera perspective model, two view geometry and computational integral imaging 3D reconstruction techniques to overcome this limitation in integral imaging systems. The image reconstruction quality of the system with unknown sensor positions is compared with conventional integral imaging with known sensor positions. We also demonstrate how the proposed method may be used to improve the image reconstruction quality even in situations where the sensor positions recorded are subject to measurement errors .