Project 4: Lightfield Camera

Overview

By using data and images from the Stanford Lightfield Camera, we are able to imitate depth of field effects - the Lightfield Camera is a 17x17 array of cameras, which are able to capture an image from slight displacements and angles over a plane orthogonal to the optical axis. Because of this we can manually produce depth refocusing and aperture adjustment by processing the images.

Depth Refocusing

As the camera moves around (keeping the same optical axis direction), objects at a distance captured by the camera only change position very little while objects closer to the camera change a lot. With this in mind, we produce a refocusing of the image by keeping the same point "in focus", shifting the images produced by each camera to align with the image taken by the center camera (where the center camera is the camera at location (8, 8)). The alignment is determined by an alpha so that shift = alpha((u, v) - (u', v')) where (u, v) is the postion of the current camera and (u', v') is the position of the center camera. alpha determines the depth of field and in this example ranges from (-0.5, 0.5).

Aperture Adjustment

We can use the previously used technique to imitate a change of aperture. Rather than using all the images taken by every camera, we only focus on some subset cluster of cameras. In order to mimic a smaller aperture, we use a smaller subset. A bigger aperture is created with a bigger subset. In this case, we use an increasingly small subset of cameras, bringing a focal area more and more into focus, from all the cameras to only the center. The alpha is kept constant at alpha = -0.1.