Engineers from the University of California, Santa Cruz, have developed some software that could change the way people search the internet and their own computers when looking for photographs.
Peyman Milanfar, a professor of electrical engineering in the Baskin School of Engineering at UCSC, and graduate student Hae Jong Seo have found a way for computers to recognise/categorise what they 'see' in an image or video. Existing technology can do this but it can only search for and distinguish individual objects in a database of images after a time-consuming training phase.
In a press release Peyman Milanfar said: "When you search Google Images, you type in a term and it gives you returns from pages that have that text in them. We want to be able to upload an image and use it as a model for finding similar images
The team have developed an algorithm which means computers can automatically recognise objects in images and actions in videos. So, if you had an image of a cat and you wanted to see what other images you had containing cats, the software will look at the image and be able to find images where cats are the main subject, where they're in the background, images took in the daylight, ones took at night etc. IAlso, it doesn't matter if the image you're using as the template or the images found in the search are sharp, clean, out of focus or noisy to the software, a cat is a cat.
"What is required is that the photos in the database be indexed and tagged with our visual feature analyser first. Then, once a query image is presented, the visual features of this can be computed on the fly and compared against those of the images in the database to return matches in order of similarity
," Peyman told ePHOTOzine. "That's just two ways to use our technology - there are a myriad other ways in which the technology can be used as such. The important thing is that we have developed a fundamental and reliable engine for comparing a given image to another for similarity in a way that mimics what a human might do.
In terms of the size of the database that would be possible to search, Peyman says this is limited
only by the specifications of the computer it's running on.
"It is only limited by the computational and storage capacities of the machines housing from the images and the information extracted from them,” said Peyman. “To make an analogy, early in the development of Google, the same scaling challenges were faced. Today, one's personal computer, along with hundreds of servers deployed by Google around the world facilitate the real-time search of the web for information.
Even though people have researched this topic of many years, Peyman says their work is as good as or better than any other algorithm out there.
Funding for this research was provided in part by the Air Force Office of Scientific Research.