ADOBE TECHNICAL REPORT Light-Field Capture by Multiplexing in the Frequency Domain

In previous works light-field capture has been analyzed in spatio-angular representation. A light-field camera samples the optical signal within a single photograph by multiplexing the 4D radiance onto the physical 2D surface of the sensor. In this paper we develop a theory of multiplexing the radiance in the frequency domain by optically mixing different spatial and angular frequency components and then capturing the signal by the sensor. We derive a mathematical method for recovering the multiplexed spatial and angular information from the frequency representation. We have developed several prototypes of those frequency domain light-field cameras. They fall into two classes: (1) based on arrays of "active" optical elements, like lenses and prisms, and (2) based on "passive" optical elements, like masks, meshes or pinholes. Both types of cameras are understood and described in a single mathematical formalism in frequency domain. We show examples of novel view synthesis and refocusing based on this frequency domain multiplexing and discuss a number of advantages of the new method.