What Is... Digital Noise?

One of the major difference between a consumer digital camera and a digital Single Lens Reflex (dSLR) is that the consumer cameras produces images with more noise when using high ISOs (sensitivity to light) and long exposure times. Professional cameras, while not immune, suffer less from these noise issues. Noise is seen as 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.]

The Digital Image

When we hear 'digital' we often think of high quality images. Digital sound does not degrade no matter how many times you play it. Digital images can be saved forever and will still print just as it did years before.

But the image in a digital camera is created by 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 basically 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'. It usually has one light sensor 'painted' on it. Generally speaking, it corresponds to one pixel in your digital image.

When photons or light strikes the image sensor, electrons are produced. These electrons 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. Heat can be from the ambient temperature around the camera, or from heavy shooting.

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.

Increasing the ISO in a digital camera is like turning up the volume on a stereo. You are increasing something called “gain” to get more of something. When your stereo increases the gain it amplifies the audio signal to get more volume. On a digital camera, increasing the gain increases the sensativity to the available light.

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”. This means that the primary audio signal increased in amplitude, but so did any background noise such as hiss.

Noise at high ISOs is similar to these audio counterparts. 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 amplification is increased to compensate for the weaker light level, the noise is inherently amplified. 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, or any electronic device with a "volume" dial.

So why does a digital SLR have lower noise than a pocket sized digicam? The photosites are larger and further apart. This has a two-fold effect. Since each photosite is further away from its neighbor the noise generated by the neighbor is further removed and is not as easily contanimated. Secondly, 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.

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 software 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 are a number of image editing software solutions 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.

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. Many consumers simply want small compact cameras.

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.

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:

• A higher ISO introduces noise.
• Noise Reduction Software will clean up the noise in some images, and sometimes it's done well enough that you can't really tell the original image had unacceptable noise level in it.
• A slow or long exposure introduces noise.
• If you must have the option of using high ISOs (e.g. to capture fast action in low-light situations), then get a dSLR. Since many amateur photographers purchase 35mm SLR cameras today (especially since they are relatively cheap), I don't see why they should not be able to use a dSLR as easily. The only roadblock to owning a dSLR for many today is its high price, though even that is steadily coming down as competition heats up among camera manufacturers. We can now obtain a dSLR for under $1000, including the lens.

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?

1. Avoid slow or long exposures.
2. Use lower ISO settings. (For more information on why see text following.)
3. When comparing digital cameras, all things being equal, favor the one with the larger image sensor. For example, Camera A may be 3MP using a 1/2.7 in. image sensor (sized at 5.3 x 4.0 mm), while Camera B may also be 3MP but uses a larger 1/1.8 in. image sensor. Camera B will probably produce images that have less noise than Camera A.
4. Purchase a noise reduction software. This will allow you to take pictures and not miss some great shots. You'll find that a good noise reduction software usually does a pretty good job of 'cleaning' up the noise to an acceptable level in most of your noisy images.

This illustration shows the approximate sizes of the most popular image sensors: