Project 4: Face Morphing

About

In this project, we experiment with depth of field effects that can be created using light field images. The image set used in this project is taken from the Stanford Light Field Archive

Depth Refocusing

A set of light field images can be obtanied by using an array of cameras to take an image in many different perspectives. The dataset used in this assignment used a 17x17 array of cameras. If we simply added all the images together, we can simulate depth of field where only distant parts of the image is in focus because it varies very little across the cameras (in contrast to close-up areas).

To achieve different areas of focus, we can, before averaging, choose a particular light field image (u, v) as a reference, find the relative shift amounts for all images (u', v'), shift by their offsets, (Dx, Dy) * c, where c is a constant to vary the distance, and then average. The result is varying distances of focus.

Aperture Adjustment

In addition, using light field data, we can simulate different aperture sizes. We can do this by choosing varying amounts of images to average given a reference point (the center is used below). By increasing the radius of the images we average, we can generate an increasing large depth of field effect on the reference location.