A manifold lifting algorithm for multi-view compressive imaging

We consider a multi-view imaging scenario where a number of cameras observe overlapping, translated subimages of a larger scene. To simplify the acquisition and encoding of these images, we propose a non-collaborative compressive sensing protocol at each camera. We discuss a prototype algorithm for joint reconstruction of the images from the ensemble of random measurements, based on the geometric manifold structure that arises from the varying camera positions. Even when the camera positions are unknown, we demonstrate that it is possible to simultaneously resolve the images and register their positions using only the random measurements.