Generalized Depth-Of-Field Light-Field Rendering

Typical light-field rendering uses a single focal plane to define the depth at which objects should appear sharp. This emulates the behavior of classical cameras. However, plenoptic cameras together with advanced light-field rendering enable depth-of-field effects that go far beyond the capabilities of conventional imaging. We present a generalized depth-of-field light-field rendering method that allows arbitrarily shaped objects to be all in focus while the surrounding fore-and background is consistently rendered out of focus based on user-defined focal plane and aperture settings. Our approach generates soft occlusion boundaries with a natural appearance which is not possible with existing techniques. It furthermore does not rely on dense depth estimation and thus allows presenting complex scenes with non-physical visual effects.