Noise - The Real Deal

If you have a chance to read the article on Noise Reduction by Joe Farace in the Oct 2009 issue of Shutterbug magazine it will enlighten you as to just what the generic term of digital noise really is.

OR

You could just read on and I will synopsize it for you. Yes it does  get a little geeky but I will keep it light.

First off there are basically five different types of noise the digial medium generates. These are dark noise, random noise, signal noise, and amplified noise. That stuff that causes technical irregular fluctuations that accompany a transmitted electrical signal but are not part of it and tend to obscure it; Digital Noise

Dark Noise - This type of noise is generated two sources, the current passing through the photo sensor cell itself as well as the individual cell lens due to its optics. It really is tinny stuff and most delicate. From the diagram below you can see the sensor sell in grey, the RGB filters, and the cell (or pixel) lens. Here the lens and sensor cell generate dark noise for heat build-up

Random Noise - This type of noise is more insidious and uncontrollable. It has to do with random (hence random noise) fluctuations of interference either electromagnetically outside the camera or within the cameras own circuitry. Fortunately this is one of the least noise issues in a camera as the manufacturer can measure to some degree the camera noise and filter it out through normal processing of images.

The last two noise factors in digital capture are the most prevalent. These being signal and amplified noise.

Signal Noise - The manufacturers have been on a megapixel war for some time now and it is coming down to the question of how many angels can you fit on the head of a needle. They would say one more. I like to think of it this way. How do you get more seats on a Regional Jet? You either make the seats smaller (which they have) or you move the seats closer together (they have done that too... Actually they have done both). Same thing holds true for digital sensors, be it a point and shoot, an APS-C sensor, or a full frame sensor. There is only so much room. If the sensor gets pushed too far you end up with the proverbial 10 pounds of poop in a 5 pound bucket. This is what has happened to the Canon G-10.  It is a fine Camera if you do not already own a G-9 as used G-9 prices rival that of the new G-10.

The new Digic IV processor from Canon does a really good job at noise control on the 5D Mk II but can only do so much when the pack 14.7 megapixels on that little sensor. This is called fixed pattern noise as can be seen in the diagram below. It is some what predictable and can be mitigated in software. This is often referred to in specifications as Signal-to-noise ratio or better yet, how well the software/sensor combination reduces this type of noise.

Amplified Noise - This comes from the sensor receiving too much information due to a high level of sensitivity or extended exposure.

The one photographers are most familiar with is noise due to high ISO. Increasing the ISO simply makes the photo sensor more sensitive to the energy being exposed to it. Compare this to a kitchen sifter, some have bigger holes (more sensitive) and others are very tight (less sensitive). The second form of this noise comes from long exposures. You can take a camera at ISO 100 and point it at a moonless sky for ten minutes and there will be noise when magnified. The APS-C sensors are very susceptible to this because of the signal noise issue. When a short exposure is made the signal noise does not have a chance to build up where as a long exposure it does (did I really need to say that... Duh!).

That sounds like a lot of tech talk so I will draw it down. Again take a look at the article in Shutterbug as Farace gets into some noise reduction software which is vital to any photographer shooting in any kind of low light. Even the newer Nikons and Canon's with unbelievably high ISO settings will generate noise at the higher levels.

Hay... Have a great weekend and I will be back on Monday... Doug