Project 1 : Hybrid Images

Paul Sastrasinh (psastras)

Summary

The goal of this project was to create Hybrid Images out of two different images. Hybrid images are images which appear to be different depending on the distance of the viewer to the image. This is done by combining the low frequencies of one image with the high frequencies of another image, so when the viewer is closer to the image, the high frequencies become visible, and when the viewer is farther away, the lower frequencies begin to dominate the higher frequencies. To seperate the frequencies, Gaussian and Laplacian pyramids are built for each image. A picture of an image's decomposition into pyramids is below.

Gaussian and Laplacian Pyramid Decompositions

Once the pyramids have been created, the images are combined by taking the low frequencies (the lowest Gaussian and lower Laplacians) of one image, and adding it to the high frequencies (the higher Laplacians) of the second image, creating a hybrid image.

Results

Results varied greatly depending on the images used and the Laplacian pyramid cutoff levels chosen. Below are some of the more successful results.