Sarah Parker (smparker)

CS 143 Image Filtering and Hybrid Images

Hybrid images take advantage of perceptual mechanisms of the human visual system in order to create dynamic images that change as a function of viewing distance. More specifically, we can exploit the fact that, though high frequencies tend to dominate perception, only low frequency image properties can be seen from a far distance. Hybrid images are created by adding two images- one filtered with a low pass filter and one filtered with a high pass filter.

Image Filtering

Creating hybrid images requires creating a high pass filtered version of one image and a low pass filtered version of a second image. The two images are then superimposed to create a hybrid. The general approach used to create the hybrid images below was to create a Gaussian filter. A cutoff frequency is determined by the standard deviation of the filter distribution. The filter is then applied to the image using the following formula:

h[m,n]=∑ g[k,r] f[m-k,n-r] where m and n are the dimensions of an mxn image and k and r are the dimensions of the filter.

This filtering process will produce a high pass filtered image. The same method is used to create the low pass filtered image, but then the filtered product is subtracted from the original image to remove the high frequency components. The standard deviation value used in the filter for the high and low frequency images was tuned separately to produce the best results. In order to preserve the image resolution, the original images are padded using reflected image content.

Hybrid Image Examples:

High Pass Filter Low Pass Filter

More Examples…