Image Quality Comparison

The most fundamental difference between the Sigma DP1 and the Canon G9 is the difference in sensor technology. The DP1 uses an APS-sized Foveon sensor, and the G9 uses a much smaller, conventional sensor. The G9's sensor uses a Bayer filter, which is used on every camera sensor except the Foveon. The Bayer filter is a mosaic of red, green and blue filters. Each sensor location on the imaging chip is covered by a single color, and each location becomes a sensor for just that color. Squares of four locations combine to form a single pixel, using two green, one red and one blue location. The Foveon effectively stacks red, green and blue sensors on top of each other, so the information for each color in a pixel is taken from exactly the same spot. The conventional system requires a bit of blurring to avoid color artifacts, but the Foveon system does not. In theory, this should lead to sharper results from the Foveon system.

In fact, it does. Things get complicated in the confluence of engineering, marketing and actual picture-taking, notably by the fact that Sigma's marketing materials triple the pixel count of DP1 images files. The Foveon has 4.6+ million pixels each in red, blue and green. Each three colors combine to form a single full-color pixel in a 4.6-megapixel image file, but Sigma calls it 14 megapixels.

Resolution
Resolution is a measure of how much fine detail an image shows. Higher resolution images have more detail, and lower resolution images are blurry. Resolution is a product of a camera's lens, its sensor, its image processing routine, and shooting conditions.

We test cameras in practically ideal conditions: to reduce vibration, we mount cameras on a heavy tripod, we set whatever shutter delay they offer, and we shoot in bright light to allow fast shutter speeds. We shoot a standard test target at a range of apertures, and pick the best result. We analyze the results with Imatest software, the finest quantitative analysis tool available for digital image testing. Imatest offers resolution results in units of line widths per picture height (lw/ph). The measurement can be directly compared between cameras of different resolutions and sensor sizes  so it's ideal for this review, which compares cameras with very different sensors.

Despite the DP1's large sensor and the theoretical advantages of the Foveon chip, the Canon G9 creates higher-resolution images than the DP1. In our tests, the G9 recorded roughly 10 percent more line-widths per picture height than the DP1.

Noise
Noise is random stuff that gets in the way of a signal. If you're talking to a friend during a subway ride, what your friend is saying is the signal, and the sound of the train is the noise. In digital photos, it is the pixel-by-pixel random variations in color and luminance away from the typical values the camera would render. It looks like grain or speckles. We shoot a GretagMacbeth test chart to test it, and measure the images with Imatest. Lower scores are better in this test.

Canon G9	Sigma DP1
ISO 100	ISO 100

The Sigma DP1 scores much, much better than the Canon G9 on our noise test  the DP1 creates only 40 percent of the noise that the G9 does at ISO 100. At ISO 800, both cameras show significantly more noise, but the DP1's score is still only 60 percent of the G9's score.

Canon G9	Sigma DP1
ISO 200	ISO 800

White Balance
We shot the Sigma DP1 and the Canon G9 in RAW mode, mostly with custom white balances. Both cameras are marketed to professionals and enthusiasts, most of whom use a RAW workflow rather than JPEGs. RAW files don't shift color data in the image based on white balance settings. They simply hold the setting as metadata, and the data is shifted in post-processing on a computer.

We got good results with both cameras using custom white balance, using white paper or a white balance target. Creating a custom white balance with either the Sigma DP1 or the Canon G9 is more cumbersome than it should be. White balance is a menu item on each camera, and setting a custom balance is a choice within the menu item. It's worth doing, and we got pretty quick at it, but really, light changes often. Photographers ought to be setting white balance often. The Canon G9 has a dedicated button for switching from single-shot to burst mode to self timer; it would have been more useful to have a custom white balance button.. To its credit, the G9 has the option of setting a custom function for one of its buttons, and users ought to set it to bring up the white balance control. The Sigma DP1 doesn t have a customizable button, but it would have been a handy touch.

Low Light
Low light shooting is typically wrapped up with high-ISO shooting. We noted a significant decrease in saturation at high ISOs for each camera, with a disastrous effect on images from the Sigma DP1. Scenes that look dull when shot at ISO 100 with the DP1 look just about monochromatic at ISO 800. The effect was most pronounced when shooting in low light. In a dark interior, we shot the DP1 wide open at 1/6 of a second at ISO 800. The result is simply gray. The Canon G9 starts out with punchier color, and retains it much better than the DP1, both as ISO rises and as shutter speeds lengthen.

Neither camera would be our pick for long time exposures, but the Canon G9 has better color performance, and at the wide angle end of its zoom range, a full stop faster lens.

Lens Performance
Which is better: being at least mediocre, but not great, at everything, or being really good at one thing, and useless at everything else? Perhaps that's the key question when comparing the DP1 and the G9. It's certainly the question about their lenses. The Canon G9 has a 7.4-44.4mm wide-to-telephoto lens with a macro function that allows frame-filling shots of objects an inch or two across. At 7.4mm, the lens also has a maximum aperture of f/2.8, allowing some available light photography. The Sigma DP1 has a 16.6mm fixed focal length wide angle lens with a maximum aperture of f/4.

The Canon G9 features image stabilization. A lens element shifts to compensate for camera movement. There are three modes: continuous, which stabilizes whenever the camera is set to shoot; shoot only, which stabilizes only the exposures; and panning, which compensates only for vertical motion. It's supposed to be useful for shots showing horizontal motion. We found the G9's stabilization effective, but the panning feature isn't useful, given the shutter lag, display lag and slow focus speed.

The Sigma DP1 does not have stabilization. Given the DP1's short focal length lens, its poor low light performance and its slow operation, it's not the camera to use in most situations that benefit from stabilization.

Our resolution test is as close as we can come to a test of lens sharpness, and Canon had a clear advantage in those results. Imatest also allows us to measure chromatic aberration. It's hard to make a lens that focuses all colors of light the same way. The differences in focus cause color fringing, or discolored edges that are most visible in high-contrast areas of images. We tested the Canon G9 and the Sigma DP1, and found that the DP1's lens is far better corrected than the G9's. The difference is stark  some of the Imatest results show the G9 has as much as 20 times the color aberration of the DP1. Sample shots bear out the result. In similar shots, the G9's color error is distracting in normal viewing, and the DP1's is undetectable.
 

Color
The Sigma DP1 should record excellent color  its most distinctive feature is its relatively large Foveon imaging chip, for which Sigma claims great things. Unfortunately, our color tests were disappointing in that respect. The DP1's colors are undersaturated significantly, and the reds, yellows and greens shift badly. Two of the reds shift to the same hue. The Canon G9 fares better, oversaturating by almost seven percent, but with less shift in color.

Canon G9

Sigma DP1

Typically, it's easier to deal with undersaturation than oversaturation in post-processing. Oversaturation causes clipping, and there is no way to recover real detail from those areas. Minor undersaturation is often amenable to adjustment in editing software. On the other hand, increasing saturation on DP1 images would have to be done selectively, because the yellows and greens are far worse off than the purples and reds. Saturation increases will significantly exacerbate the color accuracy issues.

Our charts illustrate the differences between the Sigma DP1 and the Canon G9. Imatest software creates these charts based on analysis of images of the GretagMacbeth color chart from each camera. The colorful background of the chart is a color gamut. The center of the chart indicates zero color saturation  white, all the grays, and black. Saturation increases with distance from the center, and hue varies with angular shifts around the center. Any circle centered on the chart creates a color wheel.

Canon G9	Sigma DP1

Dynamic Range
The range of values in a printed photo run from white paper to its darkest combination of ink. Really, that's not much of a gamut. On a computer screen, the range is from the red, green and blue pixels entirely shut off all the way to them glowing at full power. Again, not a big span, compared to the range from the sun's glare on white snow to a skunk's fur in shadow. It's a camera's job to translate the enormous range of light and dark in the world into the range that technology can display. Dynamic range describes how much difference there is between the brightest and darkest values a camera can capture without blowing out into pure, featureless darkness at one end and light at the other. Again, Imatest allows us to measure each camera's performance, shooting a target that shows more than 14 stops of brightness.

Sensors with large individual photoreceptors typically do better here, and the Sigma DP1 indeed outperforms the Canon G9 in this test, though the two are close at ISO 200. Despite being better than the G9, we re still a little disappointed that the DP1 isn t better. Its best score, 6.53 stops of range, is merely okay among DSLRs. It's also important to note that the G9's performance really falls off the table at ISO 800 and 1600, and is likely to limit the camera's usefulness in high-contrast, low-light shooting situations.