Evaluating and comparing DSLR image quality is very difficult, in part, due to the various imaging pipelines involved. The goal behind the camera noise comparison tool is to compare image data coming off of the imaging sensor with the minimum software manipulation applied. Software enhancements, primarily noise reduction is of concern here, can be applied to any image using a computer. Thus, the primary concern is how well the camera performs without enhancements applied.
Noise reduction is often detail destruction. As you will see in the example below and throughout the reviews on this site, noise reduction clearly reduces visible noise in an image as desired. However, there is collateral damage. Fine image details are are removed along with the noise, and sharpness is negatively impacted. Removing noise is often a positive image adjustment, but understanding the overall impact is important for tuning the adjustment for optimal final results.
Just enough is usualy the right amount of noise reduction, and the factory default noise reduction settings will typically be found too aggressive for those understanding the overall impact. For the noise test results on this site, noise reduction is turned off unless otherwise stated, such as by an "NR".
Also typically set too high is the factory default sharpness setting. Too much sharpening applied to an image causes another form of image detail destruction – halos created around contrasting pixels and aliasing (stair-stepped straight lines). Sharpening also increases the sharpness of noise, making it appear more obvious. In the noise test results, a low standard amount of sharpening, the amount I typically select for my images, is applied to the color block noise comparisons. Unless otherwise specified, a sharpness setting of "1" on a 1-10 scale is used in Canon's Digital Photo Professional (DPP) or "30" on a 1-1000 scale for Capture One processed results.
The Canon tests on this site utilize Canon's Digital Photo Professional software, free with EOS DSLR cameras. It is not the most powerful software (consider Adobe Lightroom for that title), but it is easy to use, produces good color, stores recipe information in the raw image file, and creates excellent quality TIFF and JPG conversions. Another benefit to using DPP is that the latest Canon camera models are always supported – and upgrades are always free.
Sony camera and lens tests utilize Capture One for processing. A free for Sony version of this software is available.
Manually-exposed images are captured from a Foba Gamma Studio Camera Stand-mounted Canon EF 70-200mm f/4L IS II USM Lens set to 100mm and a (usually) below-DLA (Diffraction Limited Aperture) setting of f/6.3.
RAW images are processed with a custom white balance (formerly, auto white balance was utilized), the "Standard" Picture Style or "Natural" Clarity method, and noise reduction zero'd out. RAW images are converted to 16 bit TIFF files, and Photoshop "Save for Web" is used to create the 70% quality JPG crops shown on the site.
Test lighting is from a crazy-hot 4000 watts of Photoflex Starlite tungsten lights in a pair of daylight-balanced Photoflex SilverDome softboxes. Tungsten lights were chosen for their high-quality light spectrum output.
The subject for the noise test images is a Kodak Gray Scale and Color Control Patch target, part of a much larger test target being photographed. Ideally, the camera test images will show perfectly solid colors for the various color and gray blocks in the target. Reality is that noise will be apparent, especially at higher ISO settings.
Go make some comparisons in the camera noise comparison tool. Learn how your camera performs and how it compares to others. If you can't readily pick out the difference in any comparison, it is unlikely that you will recognize the difference in real-world results.