What is noise

[jimiw]

Ive been seeing the word noise or noisy when some one is C&Cing a photo. Just what is it they are talking about? I look at the image but nothing is super obvious to me so I need a little help. Thanks Jimi

[Gbee]

In film we have film grain. The digital equivalent is noise.

Film grain could be fine or coarse, ISO50 was usually fine whereas ISO3,200 had golf ball sized grain. Grain also became obvious the more one enlarged a print, or the more one pushed a film in processing.

Noise on a digital image is gain like except it does not have the irregular pattern of film grain. It manifests itself in a number of different ways as any combination of the channels may have noise, unlike film which is compressed when processed, the digital file retains all its channels and the noise is a consequence of the signal converters in the actual camera, per channel.

Ergo a camera with the same CD or CMOS could have different noise characteristics presents.

Digital noise also follows many of the conventional rules applying to film grain. That is the higher the ISO the noisier the image will be, the boosting of an underexposed file will suddenly increase noise and long exposures will increase noise.

The long exposure is a particularly impressive subject as the tolerance for each CCD or CMOS may be wildly different, the 1D or D1x could not be used for more and 1 second for instance as it became like a snowstorm of white dots ~ and these are professional cameras of just a few years back.

Digital noise itself is merely a signal coming from the excited pixel, the pixel is moving all the time, making noise, but light excites it and it moves more rapidly, when the signal frequency matches the colour frequency the signal is strong and the noise is weak. It's really just a radio receiver, as with your radio, if you are not tuned into a radio station you only hear noise, the strong signal from the radio station overpowers the noise to give clear sound and music. The digital noise is really the same thing, get the colour to match and the noise is suppressed, the digital camera is a little different in as much as it wants to receive several clear channels at once, unlike the radio analogy which is only trying to filter out one tuned in station. Hence the noise per colour is different in a digital capture.

[Goofup]

Gbee gave you excellent info on what noise is. Sometimes a visual picture helps. This is a part of a picture of a humming bird (that obviously didn't work out as planned!), but I grabbed it to show you what noise is and isn't:



As you can see the background isn't "smooth", it looks "dotty". Usually noticable in skys, backgrounds like this, solid colors, etc. Noise reduction can usually take care of it, but you have to be careful or you'll loose some detail and things like hair start to "run together" in spots. As you can see, in my example I started to lose some detail in the humming bird's butt.

This is an example of luminance (sp?) noise- the difference between the light and dark pixels are too abrupt. There's also color noise where red, blue, etc, pixels show up.

[jimiw]

Ok now I understand, Thank you Gbee and Goofup!! It is really appreciated!!! Jimi

[jfrancho]

Excellent demonstration Goofup. Just to elaborate, your example clearly demonstrates both Chromatic (Color) Noise and Luminance Noise. Generally most noise reduction filters look to isolate the red and blue "mini goobers" and match the surrounding pixels. The Luminance Noise looks more like film grain on high speed film, and is eliminated by noise reduction filters with smoothing or blurring surface details. The term "noise" originates from "signal to noise ratio." The higher the ISO equivalent in digital, the lower the signal to noise ratio, and therefore more noise. A couple more things: by shooting at the lowest ISO possible, you reduce the appearance of noise; proper exposure also reduces noise - an ISO 200 image that is properly exposed will have less noise than an underexposed image taken at ISO 100; resizing (or resampling) to a smaller size will eliminate much of the appearance of the noise by simple attrition - you turn a 5 MP image into .5 MP image and you 1/10 the number of pixels, and 1/10 the noise.

[jimiw]

jfranco what can I say you always make things easier to understand and I do appreciate that. Jimi

[jfrancho]

No prob, Jimi. You're asking all the right questions.

[captlen]

The best explanation of noise I have seen yet. Thanks to you all.

[Tom Swaman]

Jimi,

One more point. Remember that a digital color image has two major components. One is the actual black & white layer of the image that gives the image structure and light and dark tonality without color. This is called the luminous layer or the luminous component of the image.

The other component of the image is the actual color iself or the chromatic component or chromatic layer.

Noise can affext either or both components. What Goofup has so thoroughly demonstrated is primairily luminous noise. When you see color fringes on the objects in a photo or pixels of off-normal color in relatively large quantities, this is usually called chromatic noise.

So you see, there are more than one kind of noise and the elimination or minimization of these may require some different, but distinct action on the part iof the photographer or photo editor.

Tom

[oldugly]

I knew there was more than one reason to belong to Photo Camel. Thanks to all for great explanations of "noise." Now I can explain it to others and sound like an expert.