|You are here: Home > Tutorials > What Is... Noise?|
[ Normal View ]
One of the major difference between a consumer digital camera and a digital Single Lens Reflex (DSLR) is that the former produces images with a lot of noise when using high ISOs and long exposure times, and the latter is practically noise-free (though high ISO performance varies depending on camera manufacturer and model). Noise is apparent by the presence of color speckles where there should be none. For example, instead of a blue sky, you notice faint pink, purple and other color speckles amongst the otherwise blue sky.
Just what is noise and how can we eliminate or minimize it? This article tries to explain all of this in as a non-technical way as possible. [This also means that the explanations may not be 100% technically correct. If you spot any mistake, please email me. Thanks.]
When we hear 'digital' we automatically tend to think of high quality. Digital sound does not degrade no matter how many times you play it. Digital images can be saved forever and will still print in its pristine form.
But the image in a digital camera is sourced from a non-digital component: the CCD or CMOS image sensor. Understanding how light striking the image sensor is converted into digital form makes clear what noise is and why it is present.
From Analog To Digital
An image sensor is typically comprised of a matrix of light sensors. A light sensor can be thought of as simply a device that converts light into an electric charge.
Each square of the image sensor matrix is a photosite, usually with one light sensor 'painted' on it. A photosite generally corresponds to one pixel in your digital image.
When light (photons) strike the image sensor, electrons are produced. These "photoelectrons" give rise to analog signals which are then converted into digital pixels by an Analog to Digital (A/D) Converter.
Causes of Noise
There are a number of sources of noise contamination.
Heat generated might free electrons from the image sensor itself, thus contaminating the "true" photoelectrons. These "thermal electrons" give rise to a form of noise called thermal noise or dark current.
Another type of noise is more akin to the 'grain' obtained by using a high ISO film. When we use a higher ISO, we are amplifying the signal we receive from the light photons. Unfortunately, as we amplify the signal, we also amplify the background electrical noise that is present in any electrical system.
In low light, there is not enough light for a proper exposure and the longer we allow the image sensor to collect the weak signal, the more background electrical noise it also collects. In this case the background electrical noise may be higher than the signal.
So why is using a larger image sensor better?
Each photosite itself generates electrical noise that can contaminate its neighbor. In a larger image sensor, the photosites can be physically further apart and thus be less affected by that contamination.
A larger image sensor also means that the photosite can be larger, thus have a larger light gathering capacity. It is therefore able to generate a larger signal to noise ratio.
That is why a digital camera with 6 million pixels crammed into a 1/1.8 in. image sensor has more noise (especially at high ISOs) than a 6MP digital camera using the much larger half-frame (APS-sized) image sensor.
In-Camera Reduction Of Noise
Camera manufacturers have therefore incorporated in their firmware noise reduction algorithms that kick in when a slow shutter speed and/or high ISO is used to try to reduce the noise. Depending on the quality of the algorithms, these work only to a certain extent: they do not completely remove all noise and the smoothening effect of noise reduction is frequently accomplished at the expense of fine image detail.
Noise Reduction Software
There is a number of image editing software that can be used to reduce noise in a digital image after you have taken them. Your image editing software may already have such an action, or you may download one free from the Internet. The better noise reduction software applications (such as NeatImage, Noise Ninja [review] and NoiseWare Pro) can take a long time to process one image and so may not really be a viable solution if you have lots of pictures with noise. They have their place though in a photographer's toolbox and for that one photograph that you have to take with noise or else miss an incredible shot, these software applications are your perfect noise reduction tools. In fact, no photographer should be without one.
Why Are DSLRs Practically Noise-Free?
If it has not yet occurred to you to ask it, you should. Really, why are DSLR images almost noise-free? The answer is simple: a larger image sensor!
See, with a larger image sensor, each pixel can be larger and each photosite can be a bit further away from its neighbor (of course, there is an optimum distance beyond which we'll have 'gaps'). This extra distance is often enough to prevent signal leakage from one photosite onto another -- hence much less to almost no noise!
You will find that the high ISO performance of a DSLR varies. Entry-level DSLRs are practically noise-free up to ISO 400. Better models can capture noise-free images up to ISO 1600. ISO 3200 is a stretch, even for the top line models. Turning noise reduction (NR) ON will help eliminate noise but this can happen at the expense of losing fine image detail. So far, the traditional camera manufacturers, such as Nikon and Canon, build DSLRs with better high ISO performance. Olympus, Panasonic, Pentax and Sony DSLRs seem to struggle with ISOs higher than 400. But that can change anytime.
Though the general statement that "DSLRs are practically noise-free" is true, readers should bear in mind that we have had new non-traditional camera manufacturers that have joined the DSLR bandwagon recently, and some of these relatively newer camera manufacturers do not seem to understand the importance of low noise at high ISOs and their sensors start producing noise even at low ISOs. Therefore, it pays to read the reviews carefully.
Hurrah for Bigger Image Sensors!
Why therefore do camera manufacturers not use the bigger image sensors in consumer digital cameras? A bigger image sensor means the need for a bigger lens. Unlike film that can capture light incident on it at an angle, an image sensor requires that light falls on it straight on. Bigger lens add costs, need a bigger body, etc. etc. You get the idea. You quickly end up with a camera body the size of a... DSLR.
The biggest image sensor on a prosumer digital camera is 2/3 in. sized at 8.8 x 6.6 mm (though most of them now use an improved 1/1.8 in. type). We wait for the day when an APS-sized image sensor is used in a prosumer model!
The next size down is 1/1.8 in. (sized at 7.2 x 5.3 mm) and is prevalent in most of the 5MP, 6MP and 7MP consumer digital cameras today.
Notice how camera manufacturers have 'squeezed' more megapixels into the same 1/1.8 in. image sensor. That is one reason some people say that a digital camera at a lower megapixels resolution gives images that are more noise-free than one at a higher megapixels resolution -- on the same size image sensor. More megapixels on the same sized image sensor means the pixels are closer together -- thus more noise. Of course, better in-camera noise reduction algorithms in the newer digital cameras can counter this tendency toward more noise to a certain extent. Photographers must balance the advantage of higher megapixels versus more noise (albeit reduced with the in-camera noise reduction algorithms), although camera manufacturers leave us with few choices as they all move to the higher megapixels image sensor to compete with one another.
The dSLRs have image sensors that are much larger than 2/3 in. Some dSLRs have an APS-sized (or, 'half-frame', approx. 23.7 x 15.6 mm) image sensor.
When we talk about a 'full-frame' image sensor, it is in relation to a 35mm film and is therefore sized at 36 x 24 mm. Compare these with the 2/3 in. image sensors in prosumer digital cameras sized at 8.8 x 6.6 mm, and you'd agree that the size difference is indeed substantial. No wonder dSLRs produce practically noise-free images.
Are we ever going to see bigger image sensors in prosumer digital cameras? Bigger sensors mean bigger lenses mean more expensive cameras. So that is why most of the work being done now is focused more on improving the small image sensors and writing better noise reduction algorithms. However, we believe it is inevitable for the APS-sized image sensor to eventually move down to consumer digital cameras, starting with the prosumer models.
What Can You Do?
There are a number of things to remember about noise:
Noise is a fact of life in consumer and prosumer digital cameras, and is going to stay with us for some time longer until camera manufacturers engineer better and small noise-free image sensors. Until then, what can you do to reduce the amount of noise in your digital images?
Here is a comparative illustration of the approximate sizes of the currently most popular image sensors:
|Image Sensor||Size (approx.)|
|full frame||36 x 24 mm|
|half frame (APS)||24 x 15 mm|
|2/3 in.||8.8 x 6.6 mm|
|1/1.8 in.||7.2 x 5.3 mm|
|1/2.7 in.||5.3 x 4.0 mm|
Lately there have been many "experts" that have come onto the forum boards to prove that more MP does not equate to more noise in tiny sensors and, as you would expect, they have the equations and formulas to prove it. Some contend that it is the limitations of our monitors or other technology that make it seem that way. Scientific facts are what they are. However, they miss the point. When we say that more MP in a tiny sensor results in noisier images, we are not making a scientific claim here; we are simply reporting our visual experience using the limited monitors and technology at our disposal -- and that is what matters.
Our Readers Write Back
We don't pretend to be experts in all aspects of digital photography and therefore are very happy to learn together with our readers. Their feedback to this article are published below.
From: Leo P Purcell
In Aug/Sep  I drove 4800 miles in western US and Canada with my new Panasonic FZ20. It took me quite a few days to realise that carrying the camera in the trunk in the [hot] Nevada temperatures was making a major contribution to the noise [level of the images]. Carrying the camera in the air-conditioned car interior made a significant improvement.
From: Darren Krakowski
In regards to the "What Is...Noise?" article, in which you requested additional info, I found it only really lacking in one respect, and that is with regard to noise as a function of ISO. The article eludes to the fact that noise increases as ISO increases, but does not do much to explain why. As an electrical engineer, Id like to pitch in my two cents.
Varying the ISO in a digital camera is like turning up the volume on a stereo. You increase the gain to get more of something. In a stereo, increasing the gain amplifies the audio signal, obviously resulting in greater volume. On a digital camera, increasing the gain amplifies the available light, meaning the light needs to be collected for a shorter period of time. (Of course, taking aperture into account, the shutter speed could be the same and a smaller aperture used, but the amplifying effect / gain will need to be the same, since if the aperture and shutter are changed in step with each other, the light hitting the sensor remains constant.)
With any electrical system, increasing the gain has a byproduct of inducing noise into the system. In an audio system, increased gain results in an increased signal to noise ratio, meaning that the primary audio signal increased in amplitude, but so did any background noise (typically in terms of a hiss or slightly audible static). Worse in audio is total harmonic distortion, which is clearly audible when an audio amplifier is turned up too loud and you can hear the warble, or you just notice that it starts to sound bad.
Noise at high ISOs is similar to both of these audio counterparts. Signal to noise is most easily correlated. As the gain is increased for a photosite, both the signal and the noise (present in any electrical system) is amplified. The noise at the circuit level should be fairly constant regardless of the ISO, but what has changed is the amplification. If the noise is constant but the light level is weaker, and the amplification is increased to compensate for the weaker light level, the noise is inherently amplified as well. When someone figures out how to amplify the primary signal and not the noise, it will apply to digital cameras, stereos, TV, video recorders, cell phones, you name it, and they will become very rich.
So how does this relate to a digital SLR having lower noise than a pocket sized digicam? As you correctly state, the photosites are larger and further apart. This has a two-fold effect. First, each photosite is further away from its neighbor, and the neighbor is a large part of the electrical noise generated at the circuit level. Second, a larger photosite will generate a larger signal with respect to the noise, since due to its size it inherently has a larger light gathering capacity, and thus the signal to noise ratio is greater. So, compared to a pocket sized digital camera (and to use fictitious numbers for the sake of clarity), the noise produced by the smaller digital cameras circuit might be measured at 0.1 volts. The signal produced by the circuit at the time of exposure might be 1 volt, so the signal to noise ratio is 10:1. On a digital SLR, there is less noise at the circuit level due to the increased size and distance between photosites, so lets say it is measured at 0.05. Also, the sensor can collect more light so its output might be 2 volts, producing a signal to noise ratio of 40:1, and hence a cleaner image. (This is an oversimplified explanation and the numbers are bogus, but it hopefully serves to illustrate a point.)
This also explains why low light images have a higher noise content. The signal is smaller as there is less light to gather, but the noise at the circuit level remains fairly constant. The signal is then amplified heavily to produce the image from a low light situation, the electrical noise is amplified in turn, and voila, you have a noisy final product.
Finally, remember that digital imaging noise is much like static, and thus somewhat random in nature. The noise in the above example is not a steady 0.1 or 0.05 volts, or we would easily be able to remove it. It constantly fluctuates slightly and unpredictably, and thus the need for complex noise reduction algorithms like Neat Image, which attempt to characterize the noise to remove it.
From: Joerg Colberg
Regarding your "What is noise?" article I'm
not sure I buy the explanations on why digital imaging cells produce noise.
I used to work in digital image processing (I actually processed Hubble Space
Telescope images before it got its "glasses", right after it went
into orbit) about 10 years ago, and we did lots of tests of what we called "dark
currents". Basically, we tested the behaviour of CCD chips when no signal
was coming in from a light source. There are various physical sources for the
noise (incl. temperature and quantum effects) all of which are then amplified
depending on which ISO you choose - as Darren pointed out. I am not quite convinced
about the "leaking" theory (even though I might be wrong, of course)
because if leakage caused noise you'd expect it to be very uniform, wouldn't
you? In other words, cells would leak in all directions and the effect wouldn't
really generate noise but actually de-sharpen the image.
The chip sizes, however, do affect noise quite a bit because if you have a small chip the effective size of the pixels is much smaller and this introduces larger sampling errors - noise. In a sense, the chip size translates into noise in a very similar fashion as film negative sizes translate into noise aka grain. You can't expect the same resolution from a 6x6 and a Minox negative.
From: Esquid (Nov 25, 2003)
In the responses to the "what is noise" article, Joerg Colberg expresses doubt about "current leakage" being a source of noise, and the original article also states that extra distance between photosites prevents signal leakage.
Actually, the "leakage" is a fundamental problem
with semiconductor memories - which the CCD of the camera was historically based
upon. (DRAM) memory in a computer must be "refreshed" - rewritten
- regularly to prevent it from eventually going blank. Local variations in the
manufacturing process of the CCD means that each photosite will have variations
leakage current, and the leakage changes with temperature also. Some photosites may be so bad they are "hot" pixels and are actually mapped out of the image even in a normal exposure.
See Outback Photo Noise Article for one good description of noise and the effects of leakage.
Read this article at:
All tutorial material, including images and graphics, are property of Photoxels.com
and may not be reproduced in any form without prior written consent from Photoxels.com.
Privacy Notice. Copyright © Photoxels. All rights reserved.