2003-10-08

Neat Image is, in my opinion, an indispensable tool for anyone doing serious work with digital images. Its purpose is to remove noise from digital images. This may sound odd, as hearing colors is an experience usually reserved for bad acid trips, so let me explain. Noise can appear in images as well as in sound. In sound, noise is experienced as hiss or crackle, and is not part of the audio signal being recorded; it is a distortion added by the recording circuitry. In an image, noise is visible as graininess or color speckling. It is not part of the scene being recorded, it is an image distortion added by the image recording circuitry.

The noise added to an image by a digital camera is random, but it does have statistical characteristics that allow it to be removed from the image (within certain limits) by using digital signal processing techniques. Neat Image allows you to analyze an image to determine the statistical characteristics of the image noise, and save these characteristics as a noise profile. This is a very powerful feature, because it allows you to create noise profiles that are matched to a specific type of camera or scanner. Noise cannot be removed 100% without sacrificing true image detail and giving the image an artificial, processed or painted look. But the greater the accuracy of the noise profile, the more noise can be removed from the image without removing subtle details that are part of the actual image.

I recently noise profiled my Canon EOS-1Ds digital SLR. The profiles, and the methodology used to create them, can be found here. The image below was shot with my 1Ds at ISO 400, and had NeatImage run on the right half of it using my default NeatImage settings and the 1Ds JPEG ISO 400 noise profile. Then it was processed and sharpened the way I normally handle this type of image, and enlarged 200% to better show per-pixel detail.