Real-time processing for light field imaging using an FPGA-based approach
Sven-Hendrik Voss, Marc Geh · 2020
Light field cameras can capture both angular and spatial information of light, from which depth estimation and refocusing is made possible. One main drawback of these cameras is that the instantaneous capturing is followed by complex post-processing algorithms for phase-space deconvolution to finally obtain multi-focused images. Nowadays these time consuming algorithms are implemented in software and carried out offline outside the camera itself. These cameras usually consist of a microlens array located between the main lens of the camera and the sensor. The calibration of these microlenses is one decisive step in the post-processing of the captured image. In this paper we present the newest findings on an FPGA-based on-the-fly calibration implementation and architectural concepts for efficient post-processing on the same hardware platform, targeting real-time performance of 4D video content.