Project 1 (Hybrid Images)

- Emanuel Zgraggen (ez), September 24th, 2011


The task given by this assignment is to build hybrid images using a modification of the approach described in the SIGGRAPH 2006 paper by Oliva, Torralba, and Schyns.


Hybrid Images
The algorithm takes two input images. For each of the images a laplacian pyramid with height N is constructed. The algorithm constructs a new laplacian pyramid by taking the upper K levels (high frequencies) of the first pyramid and the lower N - K levels (low frequencies) of the second pyramid. To obtain a hybrid image, this new pyramid gets collapsed back into a single image.

Image blending
Laplacian pyramids can also be used to seamlessly blend images together. The implemented algorithm follows the approach described in this paper by Burt and Adelson (Section 3.2).
The general idea is two create two laplacian pyramids LA and LB from the two input images A and B, and a gaussian pyramid GR from the binary mask R. R indicates which pixels to take from which image. A new laplacian pyramid LS is created by applying the following formula at each level of the pyramids: LS(i,j) = GR(i,j,)*LA(i,j) + (1-GR(i,j))*LB(i,j)
To obtain the blended image, we simply need to collapse LS back into a single image.


Hybrid Images

Image 1Image 2ResultGaussianLaplacianAnimation

Image blending

Image 1Image 2ResultPyramids