|
12.6
De-emphasizing the Background
In order to make the subject stand
out a bit more prominently in the scene, it’s sometimes helpful to
render the background less distracting. There are several
commonly used techniques for doing this. In this section we’ll
very briefly consider these. Keep in mind that for many images
you’ll only want to effect a very subtle change; not every image needs
to have its background blurred into oblivion. Start out with
small changes and take your time evaluating the results visually.
I strongly recommend generating different versions of the image with
different degrees of various adjustments. Save these different
versions as JPG’s and then move on to the next RAW image. Later
(preferably not the same day), you can come back and compare the
alternative versions of an image and decide which looks best to
you. Don’t be surprised if your opinion changes on this matter
from day to day. Dealing with difficult decisions like this is
simply part of the artistic process, and shouldn’t be rushed. In
time you may find that these types of decisions get easier as you
become more confident in asserting your latent artistic preferences.
The most common method for de-emphasizing the
background is to blur
it. The reason this is so common among digital photographers is
fairly obvious: images taken through an expensive prime lens with a
shallow depth-of-field (DOF) often have a “buttery smooth” bokeh, which many photographers
have come to covet and therefore wish to simulate via software.
Even photographers lucky enough to own an extremely expensive lens with
good bokeh properties will sometimes want to blur the background in
software, since not all photos will have a razor-thin DOF, and even
those that do can still have distracting background elements intruding
into the scene. Just keep in mind that the popular obsession with
smooth bokeh effects does not warrant making every one of your images
look virtually identical via massive blurring of the background.
If every bird photographer blurred the background in every image, the
effect would quickly become cliché (if it isn’t already).
Art shouldn’t be prescriptive: consider each photo as a chance to
explore the possibilities for producing something both aesthetically
pleasing and novel.
There are two common ways of blurring the background
in Photoshop. The most obvious is to apply one of the Filter > Blur tools, of which
there are many in Photoshop; the most useful of these, however, is
typically the Gaussian Blur
(though the Lens Blur can
sometimes be effective as well). In theory, all you need to do is
select the background of the image (which is often easiest by selecting
the bird and then using the Select
> Inverse menu option to invert the selection) and then
invoke the Gaussian Blur tool
with an appropriate set of blurring parameters. In practice, this
almost always leads to rather obvious artifacts, and the main
difficulty then becomes how to eliminate those artifacts—whether by
covering them up with additional processing (such as cloning them out), or by using
various selection tricks to prevent them from occurring in the first
place. We’ll explore some of the latter tricks first.
The figure below introduces the photo we’ll use for
our running example. Although the background is already fairly
blurred as shown here (despite having been acquired via a relatively
small aperture of f/11), some
“purists” might object to the distracting
pattern of leaves in the background.
Fig. 12.6.1: An
image with a naturally blurred background.
Some people would find the leaf pattern distracting, and would
thus prefer the background to be even more blurred.
In the next
figure (below) we show the same image after obliterating the background
via aggressive use of the Gaussian
Blur filter in Photoshop. The result is an image that
follows more closely the modern prescription for what is an acceptable
bird photograph: namely, a detailed foreground and a “buttery smooth” background (or “bokeh”). Whether this is more
aesthetically (or artistically) pleasing to anyone is an entirely
different matter, but for now let’s assume this is the effect you’d
like to achieve with one or more of your photos. We’ll now
proceed to dissect some of the technical issues that can arise when
employing aggressive blurring.
Fig. 12.6.2:
The same image as above, but after applying
Gaussian Blur to the background. Some photographers
prefer this look, while others consider it cliché.
The first task is obviously
to select the background. As already discussed, you can do this
either directly (by selecting the background with the Quick Selection tool, for example),
or by first selecting the foreground and then inverting the selection;
another option, if the background is already of a fairly uniform color,
is to use the Select > Color Range
option. In the figure below, the left pane shows the foreground
after selection. The right pane shows the result after inverting
this selection (to select the background), duplicating the background
to its own layer, and then applying a moderate Gaussian Blur effect. Notce
the slight brown haze around the periphery of the bird, particular
around the bird’s head. This is an extremely common—and often
very frustrating—problem when attempting to blur an image’s background.
Fig. 12.6.3:
Halo artifacts from Gaussian Blur.
Left: after selecting the foreground with the Quick
Selection tool. Right: after inverting the selection,
promoting it to a layer, and then applying Gaussian
Blur. The brown halo around the bird’s head is an
artifact arising from the inadvertent inclusion of some
brown pixels (from the bird) in the background selection.
The haze effect
shown above occurred because some brown pixels (from the bird) were
present in the otherwise green background layer prior to blurring;
those handful of brown pixels seeded the blurring algorithm around the
margin of the bird, resulting in the brown haze around the
subject. Preventing the haze therefore requires that the
background layer be entirely devoid of any pixels from the
foreground. In this case, the brown pixels were present in the
layer because the original selection of the bird (via the Quick Selection tool) wasn’t
prefectly precise; after the selection was inverted, a contour of brown
pixels remained in the background selection and were therefore copied
to the background layer.
There are several things you can do to prevent this
from happening. First, before inverting the selection, you can
expand the selection by 1 or 2 pixels, to make sure that no part of the
foreground lies beyond the selection boundary. You can then zoom
in and examine the selection boundary by eye to make sure that it
contains the entire foreground; if it doesn’t, you can either expand
the selection by another 1 or 2 pixels, or manually extend the
selection in areas where it’s needed (perhaps using the Quick Selection tool or a paint
brush in Quick Mask mode).
Once you’re sure the entire foreground is properly contained within the
selection, you can then invert the selection and promote the selection
to a layer.
The next thing you’ll want to do is to lock the transparent pixels on the
background layer (the new layer that you’ve just created from the
selection). This can be done by clicking the tiny checkerboard
icon next to the “Lock” label at the top of the Layers palette, as depicted in the
figure below.
Fig. 12.6.4: Locking transparent pixels
in a layer can help prevent some artifacts
of blur filters.
Locking the transparent pixels
will prevent the Gaussian Blur
from extending into parts of the foreground. If you fail to lock
the transparent pixels and then apply the blur effect, you’ll likely
notice that the periphery of the bird becomes somewhat blurry.
Locking the transparent pixels should prevent this.
A more frustrating problem is how to deal with edge
effects just within the boundary of the background layer. When
you expanded the foreground selection and then inverted it, you were
effectively reducing the size of the background selection. To the
extent that the background selection may have originally intruded by a
few pixels into the foreground, contracting the background selection in
this way was productive. But if you take this too far you’ll end
up with a background selection that doesn’t contain all of the
background pixels. In the figure below we show a zoomed version
of the blurred image from above. Take note of the line of
brighter green pixels along the bird’s breast and back. You’ll
also find some darker pixels that seem out of place around the bird’s
legs and feet.
Fig. 12.6.5:
Another type of edge artifact: what I call
“digital fringing”. Notice
the line of brighter pixels
along the bird’s breast and back, and the darker
pixels around the bird’s legs and toes. These are
due to the selection having been off by 1 or 2
pixels prior to applying the Gaussian Blur.
These artifacts
occurred because the background selection did not include those regions
when it was promoted to a layer and then blurred. The ideal
solution would be to make sure that the selection was perfect before
copying the area to its own layer, but in practice this can be both
difficult and tedious. A common fix (after the fact) is to use
the clone tool with a small, hard-edged brush to overwrite the “shadow” or “halo” (after unlocking the transparent
pixels) in the newly created background layer. Although
feathering the selection prior to creating the new layer can in some
instances reduce the occurrence of such artifacts, doing so is rarely
straightforward. I personally prefer to use the clone tool (or a
1-pixel paintbrush for areas of uniform color) for such tasks; with
practice, you can become surprisingly efficient at the use of these
manual painting tools. Unfortunately, if you later need to
produce a version of the image at another resolution (e.g., for
printing on large media), you’ll have to go through this painstaking
process all over again.
An alternative to the Gaussian Blur tool, for
de-emphasizing the background in a more subtle way, is to apply noise
reduction (via Filter > Noise
> Reduce Noise). Similar precautions such as
expanding/contracting the selection and locking transparent pixels are
required in this case as well, though the effect is typically more
subtle than with a Gaussian Blur,
and has the added benefit of reducing noise in the process while
possibly retaining certain prominent details (via the Preserve Details slider). I
personally avoid using this technique, since it tends to leave telltale
signs that other photographers quickly notice, though typical viewers
probably wouldn’t find anything offensive about the result. The Lens Blur filter is yet another
alternative worth exploring. Just keep in mind that a perfectly
blurred background may be what you most crave right now, but might
become less appetizing to your eyes as your tastes mature.
Remaining vigilant against cliché visual effects in digital
image processing is an elusive but important virtue for any digital
artist.
Other methods for de-emphasizing the background
include changes to the brightness and/or color (particularly
saturation) of background pixels. Depending on the properties of
the foreground, you may want to either brighten or darken (or,
alternatively, saturate or de-saturate) the background to achieve a
more striking contrast between the foreground and background elements
of the scene. As already mentioned, you should try at first to
err on the side of subtlety; a lack of subtlety is one of the first
signs of amateur art.
As demonstrated
in section 11.5, another technique for
reducing the complexity of the
background is to clone out superfluous elements such as leaves or
branches (or birdfeeders). This can be a very laborious task in
some cases, but can be extremely effective. It does require a
clone source, which needn’t necessarily come from the same image if
you’ve got other images of the same scene (or similar scenes with
consistent lighting). When cloning from other photos, be very
careful about the directionality of light; one of the strongest
indicators of artificial cloning is inconsistent shadows—i.e., shadows
indicating that the sun is both in the east and also in the west.
Finally, as we’ll see in chapter 13, it’s possible
to completely remove the image’s background and replace it with the
background from another photo. I’ve built up a small library of
background images taken from highly uniform scenes with smooth (but not
too smooth) bokeh, specifically for use as replacement
backgrounds. Replacing the entire background of an image is
typically a very time-consuming task, so I personally only apply this
technique for images that are already extremely promising. My
personal preference, however, is overwhelmingly to put in the extra
effort in the field to get the right angle that will give me a nice
background naturally, so that I don’t have to resort to tedious
post-processing techniques to artificially improve the background of
the image after the fact. Nevertheless, I often hedge my bets by
taking a few pure-background images when possible, to add to my archive
of backdrops ready to be plugged in when needed during post-processing.
|
|
|