This is the basis of our work with Facebook which can capture a light field using 6 or 24 cameras. We run a Job on cloud and it comes back as a 3D scene ORBX asset you can load into Octane for Nuke, Unity etc.
You need the camera. The software to process is hosted by us on the cloud, I don't if FB can oss it. At least the ORBX file is open and loads in Unity Nuke and AE
Also: how much processing is it required to get video light fields from video recording (with an array of cameras I suppose)? I mean: does it scale?