IBM, working with researchers in Europe, claims to have developed a better way to search online for photos and videos than current methods used by Google and Yahoo.
The developers call their technology SAPIR, for Search in Audio-Visual Content Using Peer-to-Peer Information Retrieval. It indexes and analyzes "low-level descriptors," or attributes such as color, layout, shape and sounds, in photos and videos. The technology then compares those descriptors to other existing photos to help identify what's in the picture.
That's different from the approach taken by most existing search technologies, which typically sift through images based on text tags assigned to the photos.
In a demonstration video posted on YouTube, the technology doesn't work perfectly. The video shows a man walk into a square in Madrid and take a photo of a statue. He then searches for similar images using SAPIR. He doesn't find a match so he adds the keyword "Madrid" to the search. The result is a group of photos of landmarks in Madrid, which he scrolls through until he finds a match.
With improvements, the technology could produce applications that might let someone take a picture of an item and discover stores that sell the item. Or doctors might be able to use it to assist with diagnoses, IBM said.
SAPIR uses peer-to-peer technology, so there's no central point of failure, the company said.
The researchers are still tweaking the technology and it will be "some time" before it's productised by IBM, Ari Fishkind, an IBM spokesman, said. Anyone who participated in the development of the technology is free to incorporate it into products, he said.
A demo of the technology that anyone can try is available.
SAPIR was created by researchers from IBM in Israel; Max-Planck Institute in Germany; Eurix, the Institute of Information and Science and Technology and the University of Padua in Italy; Xerox in France; Masarvkova University in Czech Republic; Telefonica's research arm in Spain; and Norway's Telenor.
SAPIR isn't the only search development project based in Europe. The Quaero project, once called a "Google killer" by former French president Jacques Chirac, first came to light in early 2006 and is now funded in part by the French government. It was initially a joint France-Germany project, but the two countries have split and now Germany is funding its own search research project called Theseus.
In addition, the main search providers and some smaller companies are also working on ways to improve image search. For example, companies including Eyealike and LTU Technologies have developed image-matching technologies.