Recent advances in digital still cameras (DSCs) are nothing short of amazing: resolutions increase at a steady pace, point-and-shoot cameras fit easily into a shirt pocket and cameras with high-end features are becoming very affordable. While it might seem like there’s not tremendous room left for improvement, consumers will be amazed at what’s in store for the rest of this year and the beginning of next year. Much of this will be made possible by the digital media processors that are at the heart of such cameras.
Digital media processors play a critical role in the DSC market, by providing original development manufacturers (ODMs) the ability to develop product lines with multiple advanced features, such as noise filtering for low-light environments, image stabilization, face detection and tracking and high-definition (HD) video, without compromising camera performance. The benefit of digital media processors is that they are highly flexible and allow processing within the camera to occur in real time with no delays.
One thing that has particularly bothered users of DSCs is their performance in low-light conditions or where there is fast action. In low light conditions, slower shutter speeds have to be used to obtain proper exposure. This causes blur from both camera motion and subject motion, and this is made even worse by the use of long focal length zoom lenses. Sporting events are often held either in the evenings or indoors where lighting isn’t optimum, so users can encounter both problems simultaneously when trying to capture their son scoring the winning touchdown or their daughter shooting the winning basket.
Camera and sensor manufacturers are working hard to improve sensitivity in the CCD sensor so that photos have a much higher quality in such conditions. Semiconductor companies are developing advanced noise filter algorithms that remove noise yet preserves detail. Until now, most cameras have had a maximum ISO sensitivity of around 400, and camera makers are now achieving sensitivities of 800 or 1600 and more. Not just jumping by a factor of two or four, the goal is to attain an ISO of 10,000 or more.
Subtle Differences
In performing the necessary signal processing, engineers take advantage of human perception and our visual expectations. For instance, as lighting goes down, our eye perceives less color. Thus, an algorithm can trade off some color saturation to get visual clarity and we don’t notice the difference, but still capture the moment.
Image stabilization is also a very desirable feature when a telephoto lens emphasizes any camera shake. Manufacturers have been working on optical stabilization methods, but it’s less expensive to implement digital techniques. A sophisticated algorithm, for instance, can determine the direction of the blur and compensate for it. Compact cameras with this feature are already hitting the market this year.
One tradeoff with faster performance, as measured by the time from when you press the shutter button to the image being captured or the time between pictures, is that cameras must move more data more quickly, especially as the number of pixels increases. Today’s digital media processors are powerful enough to process all the data, without slowing down. Instead it is the autofocus time and sensor readout time. Today’s CCD sensors for a 10 or 12-megapixel camera need roughly half a second to read out the data, which means pictures at two frames per second. However, developers are working on CMOS sensors that are faster. Earlier this year a system was demonstrated at the PMA trade show that consisted of an 8MP CMOS sensor from Micron coupled with a TI digital signal processors based on DaVinci technology that read out and processed image data at five frames per second. Sony has demonstrated a 6MP sensor with 30fps read out speeds. Such sensors could soon move into mainstream point-and-shoot cameras.
Power Surge
Higher resolution photos require more powerful image processing, which also means more power consumption. Actually, there are three main contributors to battery consumption: the signal-processing chip, the image sensor and the LCD viewing display. In each, area efficiency is improving so that while first-generation small/thin cameras could hold a battery that allowed 100 shots per battery charge, doubling that to 200 shots is becoming the norm, and already some cameras are achieving 300 shots or more per charge.
DSCs are also starting to adapt to a wider variety of scenes. For example, one new feature to look for includes face detection and tracking. Using sophisticated algorithms, a camera can detect faces in an image and automatically optimize settings for focus and color balance. Other post-processing options can correct dark shadows or bright highlights that might otherwise degrade an image.
HD Video
Another highlight is the ability of point-and-shoot cameras to capture HD video. For instance, Kodak’s recently announced EasyShare Z1275 is a 12MPcamera with 5x optical zoom. It can also capture video in the MPEG-4 format (720p30 – 720 lines, progressive, 30 frames per second). A one gigabyte memory card holds between half an hour and an hour of TV quality video or 8 to 15 minutes of 720p HD video.
Consumers are also looking for new ways to view images, especially now that they’re being captured in HD. To accommodate the differing video standards, a processor—whether in the camera, a docking station, or in a card reader—will allow high resolution still images and HD video to be displayed in full resolution on a HDTV.
While we don’t miss the chemical processing of film and enjoy the immediacy of digital images, there is still something comforting about pulling out a shoebox and looking at prints together. Creative companies are coming up with new display options, and one that’s exploding in popularity is the photo frame where an LCD with memory can store images and run a slide show.
Hide & Seek…Finally
Meanwhile, our cameras and hard disks are packed with thousands of images, and finding and viewing the ones we want isn’t always easy or convenient. To help you find photos, sophisticated signal-processing algorithms will make a difference, possibly in the next two to three years. Using face-recognition software, you will be able to point to a face in one image and ask the software to find all other images with that same face. Also note that besides raw image data, digital photos also include metadata such as the date, time, camera specific information, and soon perhaps even location and subject recognition information. Thus, with intelligent software you could search for a given person along with a date, and for instance, find all the photos of your daughter at her birthday party.
As GPS becomes more popular, it is a function that is now showing up in camera phones and will allow the user the possibility to see location data in the metadata. Combined with date information, it will become a cinch to find all the photos from your Caribbean vacation last year. There is an immense amount of data that could be embedded in an image with the proper chips, and the possibilities are endless. As always, the consumer will drive such developments.
Cam Phone Improvements
Everyone knows, too, that a camera in a cellphone is no longer a rarity, and in the phone imaging capabilities are also improving dramatically. However, signal processing plays an especially key role because these embedded cameras typically have a small sensor and a tiny lens, so the raw image quality isn’t great. Today the median resolution is 2MP, and next year the norm will move to 5MP or even 8MP. Further, all the other issues already mentioned apply to phone cameras, but on an exaggerated scale. Consider low-light issues, where if a phone has a flash at all it’s a tiny LED, so image-enhancement algorithms are especially useful in this class of camera.
Cell phones are expanding the scope of imaging; people now take photos at the spur of the moment when they normally wouldn’t have done so before. The cam phone is evidence that affordable cameras are breaking up into different segments. In years past, because of prices, there was typically one digital camera per household. Today, though, in addition to multiple cell-phone cameras, a family might have a high-resolution DSC/video camera for important events, a waterproof pocket camera for the teens and a high-zoom camera specifically for sporting events.
Clearly, to implement all the features noted in this article a camera needs a chip that does more than control the sensor and move data into memory. Digital processor horsepower is increasing to handle ever higher resolutions and advanced processing tasks, and engineers are hard at work developing the algorithms that makes it all possible. Consumers will be amazed at what cameras coming to market as soon as the end of this year will be able to do. yy
Peter Labaziewicz is Chief Technology Officer, Digital Imaging Business Unit, Texas Instruments.
