Noise and the Digital Image

sorcadmin's picture
All Photos © George Schaub, All RIghts Reserved

This is the original image shot with a 4MP digicam in 2001. On the face of it the image looks fine, a Montana sunset shot at ISO 200.

Noise 2

When I enlarge the image the noise comes screaming through. This noise is most apparent in areas of continuous tone and where colors "meet." (Noise2)

Noise detail

To help you see the noise better I have enlarged it even more and added a bit of contrast (Noise detail.)

When a noise reduction algorithm is applied post exposure the "grain" smudges out and becomes less apparent, but at the expense of sharpness. Yes, you can reduce noise, but at what price?

The price you pay is sharpness in image details. Note the soft edge picked up by the mountain silhouette.

The way to keep some edge on the silhouette is to dupe the image on a new layer, apply noise reduction and then use a layer mask to remove the effect of the noise reduction layer where the details come through. This can work fine for sky where a certain softness can be tolerated, but often looks false, and certainly does not help when you want to preserve details. If not done with precision you get a smooth and mottled image in one, not a very good result.

Noise reduction has improved considerably in newer cameras thanks to greater light gathering efficiency and improved NR software in the camera's image processor. But older images still present a noise problem.

Share | |