Pittsburgh (PA) – Graduate students at Carnegie Mellon University have developed two tools that leverage the vast image resources available on the Internet. Their tools allow users to quickly and easily create custom, professional looking images, and with little technical know how. The application has grown out of scientific research designed to analyze qualities found in images for the purposes of tagging them with meta data.
While the system is still not as robust or capable as a true Photoshop-like rendering, users with almost no skill or training can author projects very quickly. It only takes a few clicks to take a source image like this:
and digitally enhance it into an image like this:
The first tool is called Photo Clip Art (PCA). PCA connects to a website called LabelMe which contains thousands of available clip art images. The big differences between traditional clip art placement and LabelMe is the way in which the photographic data is categorized. Their tool automatically examines the source image to determine things like camera angle, lighting, scale, etc. Those meta tags have come out of research projects developed at Carnegie Mellon, those designed to take a photo and digitally analyze it’s geometry and attributes efficiently and completely.
One the meta tags are identified, a type of point-and-click operation begins whereby the author can select images at will. Placing them on the photo is as easy as drag and drop. Automatic blending enhancements are made with others available. With little or no experience a source image of questionable worth can be transformed into a completed product robust in all its demeanor.
The other tool they’ve developed is called Scene Completion. This tool is something many users will find most desirable. It allows a particular scene to be edited, augmented, enhanced or changed in some way. For example, suppose you have an otherwise lovely photograph. But, unfortunately, it has a finger shown in the lower-right, or there’s a truck right where there shouldn’t be one. This tool will search through the many millions of images available at sites like Flickr to find candidate photographic elements which can be used to “fill in the gaps” with your scene.
Unsightly trucks, the occasional finger, entire buildings, all of it can be removed (or added) as necessary. And all thanks to the advanced research projects which allow computers to analyze photographic geometry into searchable critera.
Projects like this are continuing to grow in complexity and scale. What was not possible a few years ago is now becoming possible. The idea of Photoswapping in this way may coin a new phrase. And while the ideas envisioned today by the researchers, even those things which are not possible today, they will soon be possible through simultaneous hardware advances like multi-core computing and larger memory capacity.
As the cost of hardware resources comes down tremendously, abilities will be exposed to more people allowing possibilities we never even dreamed feasible to simply be run of the mill.