Developing an HDR 3D imaging algorithm for an ultra-thin light-field camera

Abstract

Due to the limitations of camera sensors, there are situations where a single shot does not adequately capture the brightness across all areas of a scene. These sensor limitations can be overcome through high dynamic range imaging technology. Bracketing, which involves capturing a scene multiple times with different exposure settings, is not suitable for capturing moving objects and can be time-consuming. In this study, we propose a compact light field camera that can capture images with different exposures in a single shot. Light field cameras come in two main forms: the first uses an array of multiple cameras, and the second places a micro-lens array behind a large lens. The ultra-thin light field camera developed in this study is designed with a micro-lens array attached to the sensor, with each lens having a different aperture size to capture different exposure results in a single shot. After preprocessing the captured sub-aperture images to enhance image quality, we propose an HDR algorithm to obtain the image with a higher dynamic range compared to individual sub-aperture images. We suggest the equation designed based on exposure time to assign different weights based on the aperture value. This modification allows to achieve HDR imaging in a single shot. Furthermore, exploiting the disparity information from the ultra-thin light field camera, we propose stereo depth estimation process including camera calibration.