This webpage displays some of my results for the Brown University CS 1430: Introduction to Computer Vision Hybrid Images Project. Hybrid images are static images that appear to be two different images wen viewed from far away and close up. In a hybrid image, the high frequencies of one image are blended with the low frequencies of another image. The human eye blurs very high frequencies in order to avoid aliasing of the information. This means that from far away, the eye can only see the low frequencies, whereas in close proximity, the high frequencies dominate the visual allocation of resources in the visual system. This leads to the illusion that there are two different images when you are standing at different distances

I formed the images using a combination of Gaussian and Laplacian pyramids. I first constructed the pyramids for both images by using a Gaussian filter, downsampling the image by 2, and subtracting the Gaussian image from the "original" image at that level of the pyramid in order to obtain the Laplacian image for that level. I iterated through this N times. This resulted in a Gaussian pyramid of size N+1 and a Laplacian pyramid of size N. Then, I started with the smallest (blurriest) of the Gaussian images from the low frequency image on level N. With this image I upsampled, blurred and added the Laplacian of the low frequency image. I repeated these steps until hitting the first cutoff frequency. When I hit the second cutoff frequency, I started adding the Laplacian of the high frequency image. Depending on the cutoff frequencies, this sometimes led to an overlap when two Laplacian pyramids were being added. The combination of these steps resulted in the hybrid image.