1
What is digital camera noise?
Asking
somebody what noise means in digital image you can get an answer that noise is
something appearing in it that is absent in real scene. Look at the image below
captured at sunset and at its fragment taken from the upper right corner. I
doubt whether you
see something irrelevant there using zoom, say 50%.
But at zoom higher than 150% (look at zoomed fragment below) foreign noise
structure having nothing to do with sky and cloud becomes strongly visible.
There are
two main types of such noise: chrominance noise and luminance noise. The former
usually appears as small, off-colored spots or specks in the image while the
latter usually appears as small, dark spots that often look like film grain.
Both types of noise degrade the image quality and annoy a viewer, making image
unnatural. Unfortunately noise is an essential part of the Universe that is
necessarily present in any event of our life including image capture process.
Technical progress can reduce its appearance too much but cannot remove noise
completely. It is present everywhere in an image even if you capture it with
professional camera. Zoom in an image like was done above to be certain that it
is true. But doing that you can notice an amazing fact that the same noise may
annoy you differently on various image contexts. It is explained by
characteristic property of human visual perception when any foreign details may
be easily observed on uniform tonal areas (such as blue skies and shadows) while the same ones may be practically invisible on textured areas or on
edges.
2
Noise origin in digital camera
Working
with films not long ago any photographer could anticipate what noise appearance
might be in image. One can decrease noise if necessary by using fine-grained
films with low sensitivity or by special chemical processing. Dealing with
digital camera everyone is far from certain how noise can degrade an image
under given camera settings. Where is the problem? The problem is that digital imaging is a highly manipulative
medium. It means that the final color image you look at the display is the
result of the
process of converting capturing
light on a electronic sensor into the color value of the image pixels. This process includes converting photons into
electrons, amplification of the produced the electrical signal and using digital camera built-in
software for transformation of electrical signal into the final color image. Any stage of such a conversion has its own forms of
noise.
What are
the main sources of noise in a digital camera image? They can be roughly classified as
random noise (photon shot noise, dark current shot noise), fixed pattern noise
(due to variations in the manufacturing of sensors, each pixel has small
differences in sensitivity) and readout noise.
First
of all each individual detector in the
sensor referred to as a pixel (the abbreviated form of ‘picture element’) converting photons into
electrons can produce unwanted artifacts due to the haphazard way that photons
arrive at the pixel. It is so-called photon shot noise representing nature’s
fundamental limit on noise performance
in light detection systems. Its level is defined by square root of a signal.
Another type of random noise in digital camera images is dark noise. It is defined
by the electrical
signal of the sensor without incoming light. This noise can be subtracted away from a captured image
by taking a dark field image just before or after the actual image.
The resulting effect of random noise is equivalent to small statistical
variations of color values that are observed even if the same pixel would be
exposed several times by the same amount of light.
Long exposures and high temperatures cause fixed pattern
noise surrounding so-called
“hot pixels”.
This noise looks as bright
or dark points in digital images. If pictures are taken under the same
settings, the fixed pattern noise does not change significantly from picture to picture. There are some
built-in reductions technologies canceling this noise efficiently by taking picture in absence of a
signal and subtract this image from the real image.
Among other main sources of noise in digital camera images readout
noise is appearing in the process of reading of electrical charges and
converting them to an analog signal suitable
for digitization. Readout noise is important when exposures are short,
but becoming less significant as exposure increases and sky background photon
noise begins to dominate.
3
How do some digital camera characteristics and settings influence on
noise level?
3.1
Sensor geometrical size
The problem is that lots of consumers are fixated on
megapixels assuming that a camera with
more megapixels must take better pictures than one with fewer. Unfortunately it
is not true. Actually
an important camera factor
affecting the quality of the final image is the image sensor size. When it is fixed the
more megapixels sensor has the smaller individual pixel size is. In fact it increases sampling density and
spatial resolution of the image produced for a given focal length lens but increases noise level and
therefore reduces the actual light sensitivity. Under these constraints, there
is a tradeoff between spatial resolution and light sensitivity and optimal
performance can be reached with much lower megapixels than you actually have. If however you are
stickler of megapixels then keep in mind that bigger sensor gives less noisy ("grainy")
image. So try to choose digital camera having high size sensor. In this case you can get higher
actual sensitivity.
3.2
ISO settings
Camera's ISO settings describe the absolute light
sensitivity of the camera's image sensor. They are often rated as factor of 2
such as 50, 100, 200, or 400 but go as high as 800, 1600, and even 3200 on some
advanced models. The principle here is very simple – the lower ISO setting is
the slower image sensor reaction to incoming light will be. It means that a
photo at ISO 100 will take half as long as to reach the same level of exposure
as one taken at 50 under other camera’s settings being equal. Experts recommend
using a lower ISO setting when capturing overly bright scenes, since it reduces
the light sensitivity of the image sensor and a higher ISO setting when
shooting under dimmer conditions (cloudy days, indoors, etc.) since it
increases the light sensitivity of the image sensor. But everyone should know
that while increasing ISO setting digital camera processor amplifies the
electrical signal thus increasing equivalent noise level. Unfortunately high
sensitivity and noise are inseparable when one always accompanies other. For
most of consumer digital camera noise becomes visible starting with ISO 200 and
higher. It means that you can forget using high ISO settings in practice even
if they might be set formally in your camera. But compromise is always
possible when sensitivity increasing can help you capture the scene. To learn
more about actual noise level of your camera look at full size test images made
under different luminance conditions and different ISO settings. You can make
these experiments yourself or look at the relevant results at http://www.imaging-resource.com/MFR1.HTM?view=
where various digital cameras of different manufacturers were tested. If you
are not going to do so there is only one advice how to minimize digital image
noise – it is to avoid
using high ISO settings.
Some our
examples below illustrate how ISO settings affect on noise appearance.
Canon EOS 350D (sky’s
fragments)
ISO
100 ISO 800 ISO 100 ISO 400
Canon Power Shot A630 (color
patches)
(F number = 4, Exposure time
= 1/125, EV = 11)
ISO
200 ISO 400 ISO 800
Canon Power Shot A630 (color
patches)
ISO
200, EV = 10 ISO 400,
EV = 11 ISO 800, EV = 12
3.3
Long exposure
Long
exposures (usually 1-2 seconds or more) may lead to appearance of pattern of
colored dots (slightly larger than a single pixel), often referred to as
"stuck pixels" or "hot pixel" noise. But sometimes you can
get noticeable noise increasing even if your exposure speed that is originally
much faster is slowed. The example below illustrates such effect for digital
camera Olympus C740UZ capturing test image at ISO setting of 400 and nearly the
same exposure value EV of 11.
ISO = 400 (F number = 4,
Exposure time = 1/125, EV = 11)
ISO = 400 (F number = 8,
Exposure time = 1/30, EV = 11)
Useful hint here to prevent
noise appearance is to
avoid slow or long exposures.
3.4
Sensor temperature
It is known that the higher
sensor temperature is the higher noise level it creates. Fortunately normal
digital camera using does not mean its extra heating. But nevertheless keep
your camera from external excessive heat and try more often
to switch it off.
4
References and links
My brief remarks above do not
pretend to complete the picture of noise origin in digital camera and to show
how to prevent this noise increasing in the process of image capture. To
everyone who is going to learn more about it I would like to advice to read an
interesting tutorials concerning digital camera image noise at
http://www.cambridgeincolour.com/tutorials/noise.htm
http://www.cambridgeincolour.com/tutorials/noise2.htm
and informative article at http://www.ronbigelow.com/articles/noise-1/noise-1.htm
written by Ron Bigelow
To learn more about various
Digital Camera characteristics visit
http://www.imaging-resource.com/DIGCAM01.HTM
To everyone who is interested in noise reduction after image
was captured I would like to advice to read an excellent paper of Michael
Almond at http://www.michaelalmond.com/Articles/noise.htm.
It deals with 22 of popular digital noise reduction tools but Michael’s
personal 5 hits by February 2005 are the following
- Noise Ninja
- Neat Image
- Picture Cooler (PC-only)
- Noiseware
- Grain Surgery
Some
reviews about Imagenomic Noiseware, Imagenomic Noiseware 4, Neat Image Noise
Reduction, PictureCode Noise Ninja you can see at
http://www.photographyreview.com/cat/digital-photography-software/noise-reduction-software/PLS_5647crx.aspx
Personally I like to work with noise removal filters
designed by JascSoftware. In Jasc Paint Shop Pro 8 it was ISO Noise Removal
filter and in the later versions Jasc Paint Shop Pro 9, Corel Paint Shop Pro X
and XI it is called as Digital Camera
Noise Removal filter. You can read a wonderful tutorial concerning this filter
at http://www.jpkabala.com/PDF/DCNR.pdf
written by Kris Zaklika.
Good
luck.
Mikhail
ImageSkill
team
http://www.imageskill.com
