Regional selection-based pre-correction of lens aberrations for light-field displays

The display quality of lens-based integral imaging systems can be negatively affected by lens aberrations. However, different regions of an elemental image (EI) are covered by different fields of a lens, leading to variation in the aberrations within a single EI. And the pre-correction of aberrations using a Wiener filter can produce ringing artifacts in regions with severe aberrations, while the use of a convolutional neural network (CNN) for pre-correction can produce an image that is more blurry than with a Wiener filter in areas with slight aberrations. Therefore, a combined pre-correction method is proposed in the present study based on the variation in aberration severity, with each EI is divided into regions with slight aberrations and regions with severe aberrations. The former is pre-corrected using a Wiener filter, and the latter is pre-corrected using a CNN. The regions are then stitched together to produce a joint corrected EI for final optical reconstruction. We validate our approach using simulated and experimental data and compare the results to Wiener filtering and CNN learning individually.