Error analysis of depth estimations based on orientation detection in EPI-representations of 4D light fields

In the last years, interest on depth extraction from 4D light fields has increased. These techniques rely on the detection of slopes of planar structures in the light field function. These structures are also found as linear structures in epipolar image representations of the light field. Since EPI-representations are 2D signals, known orientation detection methods can be applied. This work shows that orientation estimation suffers from two types of errors: Systematic errors and random errors. A theoretical expression for the standard deviation of random errors has been formulated and verified.