Deep Color and Noise


Original 20 megapixel RAW image
Nikon D500 DSLR, ISO 100
















Image cutouts
(click to view at 100% size)
left: original image
right: edited image, noise amplified


The above photo illustrates what can happen when a photo is edited to increase brightness and contrast.

The original photo (top) has no visible noise. The measured noise in the dark background areas ranges from 0.3 to 0.6 per RGB color, based on a scale of 0-255. These areas are severely out of focus, so any variance in pixel brightness (within a very small area) can safely be attributed to noise. In terms of color depth, this image has about 9 bits of noise-free color. This very expensive DSLR produces RAW images with 12 bits per color. In this case, the lowest 3 bits are noise.

The left cutout is from the original image, and the right cutout is from an edited version of the image. Brightness was increased in the middle range. Contrast was increased using a method which emphasizes low-contrast areas and gives the eye a higher perceived brightness range.

Click on one of the cutouts to view at 100% size (a 2nd click may be needed if your browser window is smaller than the image). You can see noise in the edited image. The noise here ranges from 1.7 to 3.9 on the 0-255 scale. In terms of color depth, there are only 6-7 bits of noise-free color. Editing the image has amplified the noise.

This illustrates the advantage of images with a higher bit depth than the 8-bits provided by JPEG. A higher bit depth, if free of noise, can give you more latitude to alter the image without introducing visible noise. RAW images provide 12 or more bits of color depth. This is one reason why processing with RAW images is popular among serious photographers.

However, RAW images often have noise beyond 7 or 8 bits (esp. with small cameras), so the additional color depth may not be very useful.
The JPEG images out of the camera have been processed inside the camera to suppress noise, and have 8 bits of cleaner data. Whether you can do better than the camera JPEG depends on how well your photo editor can suppress noise while preserving detail. You are competing with the camera maker's electronics and internal software.