A Face Like No Other

A Face Like No Other


Unless you’ve been hiding on a remote island somewhere, by now you’ve heard of the new face tracking technology that’s been sweeping the industry. Some manufacturers have created their own proprietary systems while others are using technology developed by outside sources. And whether it’s called face tracking, face detection, or face information, for the most part, they all have the same purpose, which is to help consumers take better photos, in particular, of people.

“People are the subject of most photography,” says FotoNation Vice President, Eric Zarakov. “Face detection allows a camera to detect and lock onto a human face—and there are a lot of things you can do with that. How each manufacturer uses face information is up to them—some use it for focusing, exposure, and color balance. Others may use it just for focusing. We’re providing the technology from which each company can decide how they want to build upon that.” FotoNation’s Face Tracker automatically detects the position of a subject’s face and adjusts the camera’s focus and exposure. “It is most valuable in bad lighting conditions and at night, when photographers tend to underexpose their subjects,” states Zarakov, whose technology is currently being used by both Samsung and Pentax.

“Face detection, as we see it, is a way of driving business in the industry,” adds, David Troy, Senior Product Manager, Consumer Products, Electronic Imaging Products Division, Fujifilm U.S.A. “The great thing about technology is it improves from year to year; it just gets easier to use. And that’s how we see face detection fitting into things (at Fuji). Most people probably still take pictures in full auto mode because of the convenience factor,” explains Troy, “but that doesn’t always give them the best image quality. We think face detection is going to be that next big leap that’s really going to help the consumer take a better picture. The more pictures people take that are better, the more likely it is they’re going to print those pictures and use them in other ways—both printed and electronically.”

About Face

Chuck Westfall, Director/Media & Customer Relationship Camera Marketing Group/Canon U.S.A., Inc., concurs the new face detection technology takes a lot of pressure off consumers in terms of obtaining a good quality picture. “What I mean by that,” he explains, “is they don’t have to concentrate on camera controls as much as they would with a conventional auto focus system. Basically, the face detection idea means the camera automatically looks for the subject the photographer’s looking for, which a regular auto focus system doesn’t necessarily do. At the same time, face detection—at least in our cameras—goes all the way out to the very edge of the picture area, which means it sees 100% of the picture (because it’s searching for the face) as opposed to a smaller portion of the image that the conventional auto focus systems scan.”

The design of Canon’s system ensures that as long as there are one or more faces in the picture area, the image will be both properly focused and exposed. “(The new technology) is a great opportunity to revitalize interest in digital cameras, not that we’ve been suffering from lack of attention,” concludes Westfall. “There’s always something new coming along to perk up people’s interest and this is definitely going to be a popular feature for us.”

At Nikon, the technology is called Face Priority AF, and it’s available on every Coolpix model that’s been released since February 2005. “What’s great about the Face Priority AF is when you turn on our cameras, a square appears on the LCD panel with a smiley face,” states Bill Giordano, Senior Technical Manager, Nikon. “Once the camera recognizes facial details, the smiley face will hone in on the face, whether it’s a single person or multiple people in the photo. Even if you or the people move slightly, you can see the face detection on the LCD moving along on the subject’s face.

“And whether it’s a Style-, Life- or P- series camera, we have the face detection technology to go along with the in-camera red-eye fix, and one additional technology called D-Lighting. In a nutshell it fixes pictures that are underexposed,” states Giordano. Often, he notes, users elect not to activate the flash or fill flash when it’s needed. If they activate D-Lighting, the camera makes a corrected copy of the image in–camera, giving users both the original image shot and the corrected copy.

Face The Nation

“Face detection isn’t necessarily new to digital cameras. But Fuji has taken a very different approach to it,” states Fujifilm’s Troy. “Our face detection system is a hardware-driven system. It’s a new LSI processing chip, which offers a number of distinct advantages. It’s significantly faster and achieves auto focus and auto exposure in as quick as 4/100ths of a second, all with a simple touch of a button. It’s not only very easy, but also very fast.

“One of the other benefits of the system we developed is it also detects a greater number of faces. Where current systems achieve a maximum of three faces being recognized, Fuji’s technology will detect up to ten faces in the frame. It’s essentially looking for a few focal points on faces—the eyes, the ears, the chin, and nose. In the development of the cameras, we wanted to make this extremely easy to use and to give users the flexibility to decide whether or not to use the mode. On the FinePix S6000fd, it’s a one-touch button on the camera’s back,” Troy added.

So what does this new technology mean to the industry? Will better images prod consumers to print? “Consumers print only the images they consider to be their ‘best’ photos and face detection technology will increase this number of print-worthy images,” states Joellyn Gray, Director of Consumer Printing, Fujifilm. “In addition, many digital consumers are eager to be creative with their photos. Higher quality images with subjects looking their best will result in more scrapbook pages, larger prints for framing, and additional photo merchandise.”

“Face detection breeds confidence when it’s used. This is very important, because the trend today (in printing) is to upload images to a Web site for printing or use a kiosk in-store,” adds Lindsay Silverman, Senior Technical Manager, Nikon. “(Consumers) don’t want to print pictures with poor exposure, poor focus, and red-eye. Right out of the camera we’re giving them great pictures for printing. They don’t even need a computer.”

What Else Can it Do?

Some manufacturers, such as Fuji, believe the industry has only scratched the surface when it comes to the technology’s future use. “We see it as technology having legs beyond the actual feature,” says Fujifilm’s Troy. “The technology we’ve incorporated in the cameras is actually built upon the image intelligence technology that has been built into our processing systems for a number of years. We’re taking the best of what’s been created and re-developing it into new areas. Ultimately, the ability to improve image quality and the ability to identify/sort files, etc., will all be based on facial recognition.”

“Just as with any new and emerging technology in the digital age, there is hype and reality associated with face recognition,” states James Latham, Vice President, Marketing, ACD Systems. “Large companies like Fujitsu, academic institutions and researchers, and small emerging companies have taken various approaches to recognizing faces. Much of the work requires massive processing power and time, making it unsuitable for the personal computer despite advances in desktop processing power. Unfortunately,” notes Latham, “it will be some time before this type of technology makes its way into effective usage on the PC desktop. Early inclusion of this into programs merely serves to confuse and frustrate the consumer—so that approach is seriously flawed. When it does become available by extrapolation of current methods, or if there is a breakthrough to a new method, then it will become quite useful.

“However, there are some promising areas,” he notes. “Ultimately, face recognition, as well as recognition of other items (oceans, trees, forests, etc.), could mean that photographers no longer have to keyword their photos by hand; instead, it would be done automatically. Then the task of rating and culling the photos becomes much easier. Consumers would be able to easily find that exact picture using this technology and the industry would benefit by an overall lift.”

Canon is also working on ways to help consumers sort their images. The company recently introduced new products with a feature called My Category, which automatically “tags” an image as a portrait, landscape, etc. Consumers can also assign their own tags to images when the camera is in playback mode. “There’s a selection of pre-set tags that are built in, but there are also custom tags they can use to organize their images,” notes Canon’s Westfall. “Once images have been tagged, there’s all sorts of things that the camera and the computer can do to help make organizing the images easier.” For instance, My Category lets users create a slide show in the camera that looks at a specific category. Or, users can program it to make prints of any category. “The category tags definitely take the idea of classifying and categorizing the images a step further than we’ve ever done before.”

Video conferencing is another area we can expect to see this technology used,” says FotoNation’s Zarakov. “For video conferencing (using a wide-angle camera looking at a panel of people), the face tracking allows you to quickly auto zoom onto different faces without having to manually zoom in and out. It’s controlled through the application—you can track a face and click on it. In other words, it enables easier selection of viewing for video teleconferencing. It also enables what we call selective compression. In a scene like a video/teleconference, the rest of the scene will be static while the person’s face keeps changing. So rather than transmit and compress the video stream of the entire scene, you can just compress the changes that are happening in the person’s face, which makes a more efficient video stream.” FotoNation is also looking at the technology for ways to improve cell phone photos.

According to Alexander Loui, Ph.D., Project Lead, Kodak Research Labs, Eastman Kodak, the company has been working with face detection technology for some time now. “The traditional application of face technology is in the photofinishing area and the more recent application is red-eye reduction,” states Loui. As for the future, he notes, “the ability to detect faces and recognize people will have a big impact on the industry, especially in the image organization and search and retrieval areas.”

Consumers armed with digital cameras are taking more pictures than ever (and 40% of those images include people). Hence, they’ve accumulated thousands of images to the point that they are having trouble finding specific photos. “The ‘digital shoebox’ may be more of a problem than the real one, which at least allows you to physically view prints,” says Loui. “Many backups make the situation worse.” Hence, Kodak has also been working on automated album applications—digital storytelling—to help consumers create photo albums (the system can even help select images for the album.) In the professional arena, such as wedding photography, Loui says the technology will help photographers save time by sorting thousands of images—by bride, groom, and/or guests. “It’s a huge task. And anything that can help them automatically or semi-automatically (sort photos) will help them tremendously.”

Although it sounds like an item made especially for James Bond, Loui believes the future of face detection/recognition technology will be used to alter expressions on people’s faces. For instance, that family portrait where everyone but Johnny smiles; we’ll have the ability to correct his smile using face information (a facial feature can be extracted from another shot and help modify the expression). “If you can’t get a perfect shot,” concludes Loui, “the technology will help you get a good picture.” Seriously. yy