Project #5: Lightfield Camera

Overview

In this project, I implemented depth refocusing and aperture adjustment using a dataset of images taken over a regularly spaced 17x17 grid. This dataset comes from the Stanford Light Field Archive.

Depth Refocusing

When we take the average of all of the images taken over the grid, we see that the nearby objects are blurry since the their position vary largely between photos. However, the far-away objects vary in smaller distances and are thus sharper in the average image.

We can take advantage of this property to refocus at different depths by shifting the images by multiples of (u, v) and averaging them. The (u, v) is the image's offset from the center of the grid. Below are results from different scaling factors of (u, v).

c = -3

c = -2

c = -1

c = 0

Aperture Adjustment

We can also create the effect of aperture adjustment by averaging images within a specified radius of the point in focus. I chose the focus point to be the center of the image (using the technique from the previous part to shift the focus to the center) and varied the aperture radius in the range of [1, 8].

radius = 1

radius = 2

radius = 3

radius = 4

radius = 5

radius = 6

radius = 7

radius = 8

Summary

In this project, I learned about the different postprocessing techniques that are possible with lightfields. It's incredible what can be done using images and a bit of math. It's also amazing that such a cool new idea was implemented upon the camera, a technology that has been around for centuries. I'm excited for upcoming innovations in photography, which are sure to come.