This image reminded me of that optical illusion where, when you view it up close it looks like Einstein, but when you view it from far away it looks like Marilyn Monroe. I had always thought the effect was made by dropping a blurry image of Monroe onto a regular image of Einstein and taking the average pixel values, but could it have also been made by combining the frequency domains of a normal image of Monroe and a normal image of Einstein, just with parts of each domain removed? Are the two processes essentially the same? Or is combining on the frequency domain level better since you're combining before you sample again?
Staffemilyma53
This a great observation! You're correct that this effect is produced by using different frequencies from both images. Usually, this can be done by having the high frequencies of Einstein overlaid with the low frequencies of Marilyn Monroe. There's a project about this in the Computer Vision course at Berkeley. See this link for another example.
This image reminded me of that optical illusion where, when you view it up close it looks like Einstein, but when you view it from far away it looks like Marilyn Monroe. I had always thought the effect was made by dropping a blurry image of Monroe onto a regular image of Einstein and taking the average pixel values, but could it have also been made by combining the frequency domains of a normal image of Monroe and a normal image of Einstein, just with parts of each domain removed? Are the two processes essentially the same? Or is combining on the frequency domain level better since you're combining before you sample again?
This a great observation! You're correct that this effect is produced by using different frequencies from both images. Usually, this can be done by having the high frequencies of Einstein overlaid with the low frequencies of Marilyn Monroe. There's a project about this in the Computer Vision course at Berkeley. See this link for another example.