Take many photographs of a static scene using very high shutter speed (or electronic equivalent) from a moving camera, perhaps a nearly pinhole camera to achieve large depth of field so not to have to worry about focusing. Let the camera also be high resolution. This will result in dark, grainy or noisy images. The goal is to avoid motion blur despite the camera moving.
Next, combine successive images to synthesize a good image, as if the camera had been static and exposure long enough. Use computer vision techniques to find correspondences between images as the camera moved. Possibly use accelerometer and other motion hints recorded by the camera itself. Possibly present a not-quite-finished image in a way to harness the extremely powerful human vision system to "finish" the synthesis in our minds.
No comments :
Post a Comment