CS 143 Project 1

Hybrid Images with Laplacian Pyramids

Gili Kliger (gkliger)

First we construct a Gaussian and Laplacian Pyramid for each image. The Gaussian pyramid is simply a sequence of images, each of which is successively blurred using a Gaussian Filter. The Laplacian filter is made by subtracting the Gaussian blurred image from the previous blurred image in the Gaussian pyramid. Each successive level of the pyramid downsizes the image by 1/2. Once we have the pyramids we must combine them to produce the hybrid image. This is done by taking the first C images (where C is the cutoff frequency parameter) in the Laplacian pyramid of the image we wish to appear up close and the last set of images in the Laplacian of the other pyramid. To rebuild the pyramid we start with the lowest level of Gaussian of the low-frequency image and add succesive images in the Laplacian pyramids, doubling each image in size to reverse the process of building the pyramids. The final image in this new pyramid is our hybrid image.