A Swedish startup is combining software and humans to help make photos and other images more easily searchable online, raising privacy concerns as the technology eases the tracking of people across Web sites.

Traditionally, search engines analyze text surrounding an image on a Web site. So a search for ''Bill Gates'' might produce a photograph captioned with the name of the Microsoft Corp. chairman. But a search for a reporter's name might produce that same photograph if it had accompanied an article he had written.

Polar Rose AB is bringing facial-recognition technology to the mix. Its software scans everyday images for about 90 different attributes. If the software finds a match with images in a database, it concludes the two photos are of the same person.

The company, among many startups seeking to improve image search, believes its technology is noteworthy because it creates 3-D renditions of faces in images, allowing the computer to account for slight variations in angles and lighting.

Nikolaj Nyholm, the company's chief executive, said testing has shown up to 95 percent reliability with sets of 10,000 photos. But he said that as the collection grows — there are millions, perhaps billions, of photographs on the Internet — reliability diminishes because, well, many people simply look alike.

That's where humans come in. In early 2007, the company will distribute free plug-ins for Microsoft's Internet Explorer and Mozilla's Firefox browsers. People who post or view photos could add information such as names; there might be the occasional error, but enough people filling in the correct answer would make that rise to the top.

The idea is to label every face, even ones in the background, whether posted on a Web journal, a photo-sharing site like Yahoo Inc.'s Flickr or a social-networking hangout like News Corp.'s MySpace. The service won't index images on personal computers or password-protected sites.

Polar Rose plans to sell ads and premium services but won't charge for the basic use of its plug-ins or search engine, which is still in a ''beta'' test phase.

But there's still a cost: privacy.

Imagine yourself minding your own business when a tourist at Times Square snaps a picture with you walking in the background and posts it on a public site. Using a search engine like Polar Rose, your boss could easily find out you were out and about on a day you had called in sick.

Police, stalkers and spouses also could use the technology to track where people have been — for example, if someone has attended anti-war protests in multiple cities.

''I don't think we have all the answers quite yet,'' Nyholm said, adding that people went though similar debates years ago when search engines began indexing text.

''A lot of pictures have been published, and privacy has been assumed due to obscurity,'' he said. ''This will highlight the fact that there is no such thing as privacy by obscurity.''

It's not clear how well the service will work. Facial-recognition technology isn't error-free — people get tans; some occasionally wear sunglasses. And the human component will help only if a large number of people participate; many other human-assisted search engines have produced lackluster results.

Simon Davies, director of Privacy International, said that regardless of the service's effectiveness, technologies such as Polar Rose underscore the need for a global debate on whether to place limits on what search engines can index and to give individuals greater say.

Without such dialogue, he said, ''these technologies will keep drilling into information to create search dimensions which are infinitely more powerful than we could ever imagine.''

And he rejects Nyholm's contention that just because an image is accessible, it's fair game.

Whenever information becomes easier to find and access, ''a whole raft of new privacy issues are always created,'' he said. ''When people place their photographs on the Internet, they do not expect them to be searchable.''