DIGITAL CAMERAS: Some tech specs are more hype than help
Digital cameras arrived on the consumer market about ten years ago and have greatly improved in image quality and overall performance since then.
Digital cameras arrived on the consumer market about ten years ago and have greatly improved in image quality and overall performance since then. Capabilities for testing and measuring digital-camera performance have also developed significantly in that time. But, particularly in the consumer market, the mixing of marketing and engineering goals can sometimes provide information in a way that is more confusing than helpful.
Images produced by early digital cameras, released in 1995, tended to be noisy, dark, and short on dynamic range, according to comments by Dietmar Wueller, CEO of Image Engineering (Frechen, Germany) during an invited talk at the annual SPIE (International Society for Optical Engineering) Electronic Imaging conference (Jan. 15-19; San Jose, CA). But in cameras released just one year later, the noise was pretty much gone, sharpening was available, and overall quality had improved to the point that text captured within an image could actually be read, added Wueller, who also participated in a demonstration session at the meeting.
Ten years later, as digital cameras have moved into the single-lens-reflex (SLR) market, newly emerging mobile-phone cameras offer image quality similar to the first digital still cameras. “So expect the same developments as with still cameras,” he said.
But some of the information presented to consumers turns out to be confusing, if not misleading, Wueller said. In a number of cases, “technical” specifications offered to describe digital-camera capabilities have either been invented or rephrased for marketing rather than technical purposes. For example, “digital zoom” simply crops an image and enlarges the section, reducing the information carried in the image (as opposed to optical zoom, which maintains a full-size image).
A more recently introduced technical-sounding term that actually serves marketing purposes is software-based “digital-image stabilization,” which corrects for motion artifacts or noise in an image electronically, after the fact. Digital image stabilization is just another word for increased amplification, which creates noise in the image that then has to be electronically removed. In contrast, hardware-based optical image stabilization enables handheld photography at slower shutter speeds than normal through the use of a moving prism or gyroscopic element in the camera that help to stabilize the projected image while the photograph is being taken.
Standardized test charts, either transparent (top) or reflective (bottom), help to provide a high degree of repeatability in camera test and measurement.
Even among technical specifications that make sense, some are marketed in ways that lack clarity. One of the Fuji Super CCD digital cameras, for instance, is marketed as having 12.3 million effective pixels. In reality, though, the 12.3 megapixels in the Fuji camera consist of 6.17 million so-called S-pixels, and 6.17 million R-pixels. The S-pixels are more sensitive to bring out the shadows, while the less-sensitive R-pixels capture the highlights. The two in combination significantly enhance dynamic range, but the resolution is still 6 megapixels.
Similarly, a Foveon camera is marketed in Germany as having 10.2 million pixels, actually corresponding to 3.4 megapixels each in red, blue, and green, Wueller said. The outstanding feature of the camera is not its spatial resolution as the 10.2-million-pixel specification implies, but the vertical stacking of red, blue, and green pixels to provide direct color detection at every point on the image (similar to film). So it eliminates the need for the camera to interpolate color across adjacent pixels. A good 6-megapixel camera has much better resolution than the 3.4-megapixel Foveon camera, but may not perform as well in terms of eliminating color artifacts, Wueller said.
“So we need tests,” he said. “Because we do not know camera performance by looking at technical specs.”
Testing and human factors
A decade ago, image quality was evaluated primarily through visual analysis of images taken of test charts and of natural scenes. And ISO (International Organization for Standardization) standards on test procedures that have evolved since have facilitated development of objective and reproducible measurement methods for camera characteristics such as dynamic range, speed, resolution and noise. A white paper that comprehensively discusses test methods can be downloaded for free from the Image Engineering Web site (see figure).1 Wueller highlighted some of these tests during his talk, along with factors to consider in performing them.
For instance, in tests based on visual analysis, using natural scenes rather than graphic patterns may provide less repeatability because the lighting of natural scenes differs throughout the day and year. Testing cameras without taking pictures of natural scenes, however, may miss important system variables. For instance, a particular camera may work well in the studio, but not in very bright light. “So you must shoot many different situations,” he said.
Human factors present another important dimension in camera test and measurement, he said. For instance, “even when using ISO test factors, where do you draw the line for limiting resolution? No artifacts? Loss of detail?” Different observers may differ in evaluating resolution in the same image by as much as 50%, he said. “Image pleasantness” is another observer-relative factor that is essential for camera evaluation, but that cannot be objectively measured. “Images must be shown to several people to get impressions, which is very intensive work.”
Hassaun A. Jones-Bey