Thanks to Adobe's Photoshop picture manipulation tool, we now have more Bigfoot, space alien, and dinosaur sightings than ever before. In fact, it's hard to believe that we can't just walk out the front door and see unicorns roaming freely in the streets or Elvis crusing by in a Cadillac. Those image manipulations are going to get a lot more interesting, thanks to some researchers from Carnegie Mellon University.
The researchers are adding a completely new dimension to photo editing that goes beyond merely moving things around or cropping objects or people into the image. Bring on the third dimension. That's right. editors can now grab objects in an image and flip them around as if they were 3D models, making it possible to see the objects from different angles. Flipping a car over, for example, would allow us to see its underside, even though we couldn't see it in the original image.
Need more cabs? The tool allows objects to be multiplied, flipped, or turned in almost any way.
(Source: Carnegie Mellon)
Think of it as manipulating a 2D object as if it were 3D all along. The magic behind the software is that it scours the Internet for repositories of 3D images that generally match the object being manipulated. Those 3D images are fitted to the object's geometry in the original picture. Then the colors, textures, and lighting are applied, giving the new image a realistic look while matching the original content.
Though the software is meant to be used with digital content, the researchers found they could manipulate images in historical photos and even paintings. If that wasn't enough, those objects and images can also be animated. This was demonstrated using a handheld origami bird that could flap its wings and fly away. The same was done for a taxicab that levitated and flipped over on to its back.
The software is great at manipulating objects that have solid and rigid surfaces, but it's also great for objects with deformed surfaces, such as backpacks and clothing. Such objects are manipulated semi-automatically; the software tries its best to align the geometries while preserving the symmetries. It then estimates the environmental illumination in the image and applies it with an estimation of the objects appearance on its hidden sides (using the viewable sides of the object) to complete the 3D model. If the software can't extrapolate the data to complete the 3D image, it approximates the data using repositories of 3D models. Even though 3D image stock is limited, the repositories will undoubtedly grow due to the widespread use of 3D scanners and printers.
It's anyone's guess when the software will become available, but you can expect to see more believable aliens and unicorns when it does.