Digital cameras have leaped bounds and barriers over the
last few years. It is not uncommon for small pocket-sized
cameras to have 10 megapixels of resolution. Tremendous amounts
of detail can now be captured with a camera the size of a
deck of cards. These small cameras can really pack a punch
for image quality in good light, but may experience a common
problem known as “noise” — especially in poorly lit conditions
or dark areas of a photo. In poorly lit conditions, an increased
sensitivity to light (ISO setting), may be needed for a faster
shutter speed to avoid a blurred photo. When increasing
the gain (ISO), “noise” also increases and can appear as
grain or coloured mottling in an image, usually in the darker
portions of a picture or in solid colours. It can turn a
precious captured moment into a mottled mess if bad enough.
There are programs capable of “cleaning up” noisy photos
without sacrificing a great amount of detail. These are called “Noiseware” programs.
In this article we will look at 4 noise reducing programs
that are all readily available online. They are: Noiseware
Community, Neat Image, Helicon Noise Filter, and Picture
Cooler.
To do a comparison of the 4 programs, we will use the same
photo — a wildlife picture of a Spotted Towhee that was shot
at ISO 3200, under poor lighting conditions. It is predictably,
quite “noisy”.
This will be our sample image:
Seen at regular viewing size this
image is not too bad, but not great either for noise levels.
There isn’t much “chroma” noise (which appears as coloured
blotching and mottling), but mostly “luminance” noise which
is more like film grain. This is probably because the sample
image is from a dSLR, where chroma noise is better controlled
than on smaller digital cameras. When we blow the image up
to 100%, it’s a different story - the noise is quite noticeable
with quite a bit of visible grain:
Now we’ll see how these 4 programs
can deal with this noise and produce a smoother, cleaner
result with minimal loss of detail.
This program comes in a free
version, which we’re using, and also a pro version which
can be purchased.
“Open” the folder and select the noisy image. Use the zoom
tool to view at 100%. We make sure the preview tab is highlighted
for back and forth viewing of the original and adjustments
that are made (hold your mouse over the image left clicking
and releasing to see the difference). It should look like
this after zooming in to 100%:
With the picture loaded into
the program , the menu is presented with various
sliders for making the desired adjustments. The “settings” tab
has a drop down arrow which will show options for different
types of photos that one may want to de-noise (see next
photo below).
The most obvious are the first four, default, landscape,
night scene, and portrait. Any of these can be chosen as
to which is closest to the type of photo that is being worked
on.
Also at the bottom,
is an option called “custom”, which we used here. Another
tab at the bottom of the “noise filter” section is “more controls” This
opens up more sliders to fine tune the image at different frequencies,
but is only available to use when the full version is purchased.
To start with, the sliders
are at default levels. A quick fix at these default settings
can be achieved just by clicking the “Go” tab, and the image
will be processed as to the default levels. The image on
the left is the result.
When using the default settings, the result has
too much noise suppression. The image has taken on a “plastic” look,
and while all the noise has been removed, most of the fine
detail has also been eliminated.
Further adjustments are needed.
Using the sliders to the left, “noise suppression” was changed
from 100% to 0%. Then the luminance setting was increased to
45% which has reduced the grainy look without losing too much
detail. The colour slider was set to 35%, and a small increase
on the “noise level adjustment” sliders boosting the luminance
to 5%, The colour was backed off to -15%, and then a +1 boost
in sharpness. The image on the right is the result.
Viewed at 100%, we see the noise
has been effectively reduced. Some of the fine detail is obviously
lost, but when viewed at a normal size, it is quite minimal.
Here’s how the end result looks at a normal viewing size:
The image has a smoother look
to it, but the bird hasn’t taken on a “plastic” look either.
The background is noticeably cleaner, and is not as grainy
as the original.
As with other noise filter programs, when adjusting the
filtering in each photo, it is advisable, as noiseware designers
recommend, to play with the sliders and get a feel of how
they work, and thus achieving a result that’s acceptable.
Now the image is saved to the originating folder. The “de-noised” image
will be a duplicated separate file with “filtered” tagged
to it, and the original photo will be left intact. Note:
this is a feature of all the programs we looked at. When
saving the adjusted picture, a filtered duplicate is created,
and the original photo is left untouched.
The second program we looked
at is Neat Image. This program has a free version
with an option to buy an upgrade for more features. There
are a few steps needed to operate this program compared to
the previous, but they are easily learned.
As seen here, the program reads some of the exif data embedded
in the photo stating the camera make and model, ISO setting,
and shutter speed.
Next we click on “device noise
profile”. This will open another tab right below it which says “auto
profile”:
This is to let the program
profile the noise in the photo. So we click on this and the
image is automatically “profiled” for the amount of noise
it contains. The profiled area used is in the little blue
box. This area can be dragged and moved if a different area
of the photo is desired for profiling the noise.
After profiling the noise, we select the “Noise Filter Settings” tab.
To the right of the image are the sliders for adjustment
similar to those in the Noiseware Community program.
The little trashcan at the
top is selected to reset the adjustment sliders to default
levels. This is for starting from scratch after one image
is completed, and a new image is selected.
Then we click on the “preview” tab, and the window reveals
a rectangular area in which the filter is being applied and
is the reference area that will show the ongoing adjustments
as they’re being made.
This rectangular box can also
be dragged around the image to a location desired to use
for previewing the adjustments. As in Noiseware Community,
Neat Image also uses a default level of noise suppression
that is too strong for what we’re trying to achieve. So on
to the adjusting sliders. Taking the same approach we did
in the previous program, “noise reduction” amounts were
set to zero, and adjusted from there, finally giving the
image a 10% boost in sharpening.
The selected area has filtered out the noise quite effectively.
The background is much smoother, and the grain in the bird’s
feathers is not as objectionable.
Note: when using the sharpening tool
in noise reduction programs, it is best not to get too
carried away with large amounts of sharpening. Otherwise,
the image can end up with artefacts such as “halos” or “jaggies” around
the defined lines in the photo.
Clicking on the “output image” tab is next — the screen
will go blank — then click on the “apply” tab. This processes
the image, and it will then appear again in the window in
its processed state.
The image can now be saved to a folder with the tagged
name on it, which appears in the “output image” section on
the right. There’s also an option to “queue” in the “queued
output image” section if working on more than one photo,
and then being able to save them all at the same time.
The final result using Neat Image when viewed at a normal
size:
The background is smoother with
still some grain visible, but not objectionably so, and the
bird’s feather detail, although reduced a bit, is still acceptable.
As in the other programs,
select the image and zoom in to 100%. The window and image
looks like the screen capture on the right.
This window reveals the thumbnails on the right of the screen
from the folder from which the image was acquired. Clicking
on the “noise” tab just right of the “source” tab processes
the image to default settings. A windows appears with simplified
setting for noise reduction limited to one slider.
To be able to do more adjustments,
click on the “simple/expert preview” icon at the top in the
tool bar. This will open up multiple adjustment options,
as seen on the left.
As in the previous 2 programs, the noise suppression at
default is a little too high. The image has again, a “plastic” look.
At default, the “method” section is set to “aggressive”.
We changed this to “standard”. This immediately reduced the
noise suppression to a much lower level.
This program has a wealth of
adjustments available. To keep things simple and for comparison,
we made adjustments similar to the other programs.
The noise level at default is set at “130”. As this slider
is adjusted, the “highlights” and “shadows” sliders move
automatically with the noise level slider. The noise level
was set to 185 – this automatically changed the highlights
and shadows settings.
Note: after adjusting the noise level
slider, the highlights and shadows can be moved independently
on their own if so desired. Radius was adjusted to “8”. The
chroma noise slider was left at default.
As we can see, the background has a smoother texture, and
the feather detail has been fairly well preserved. Here’s
the image at regular viewing size:
Note: An extra feature using the Helicon
Filter, is when clicking on the individual adjustments,
a small box at the bottom of the program window explains
what each adjustment is achieving, informing the user the
good and bad points of each adjustment.
This again is a free program
to use, with the option of purchasing the fullversion.
We open the desired image and reset the Navigator Zoom slider to
100%.
Picture Cooler differs slightly from the other noise reduction
programs. When using the program for the first time, the
default noise reduction setting is already being applied.
The program it seems, automatically
profiles the noise when the image first opens, and subsequently,
saves the settings and adjustments when closed. To achieve
a starting point that looks like the original image, we move
the “Input image de-noised image” slider all the way to the
left. This reduces all applied settings showing the original
input image. This one slider adjusts 2 settings — how much
de-noising is applied to the image (right part of slider),
and the input image (left part of slider). Both values change
as this one slider is adjusted.
Now we can see all the noise
this image has, and can start adjusting from here.
The “input image” slider was adjusted to 40%. This also
adjusts the “de-noised image” setting automatically to 60%.
The luminance slider was set to 0.4 and then the “refocus
level” was set a little to the right to retain some of the
sharpness. The “refocus jaggies” slider was then moved a
little to the right to smooth out the jaggies which were
slightly noticeable. This is the result.
The background is much smoother,
and the bird has lost little detail. This is how the final
image looks at regular viewing size:
One final note about Picture
Cooler — there is also an “advanced features” tab that can
be clicked on at the bottom of the adjustment area. This
will reveal another set of adjustment sliders: these can
be used to further adjust the image if so desired.
Conclusion
All 4 programs we’ve looked at have achieved decent results
producing an image with a smoother look and little loss of
detail, albeit, with needed adjustments, and not applying
default settings. At default, all of the programs seemed
too aggressive in reducing unwanted noise resulting in a
plastic looking image.
Each program gives the option to start from scratch without
any noise reduction being applied, and letting the user decide
how much de-noising is required via the use of the adjustable
sliders.
Workflow for all 4 programs was quick and easily learned.
The only recommendation we could make for these 4 programs,
is to give each a trial run to see which one meets an individual’s
needs in respect to workflow and end results.
is
part of the 
™ Network
