| Post-processing:
a thought experiment
Imagine a beautiful
beach spoiled only by a discarded Coke can on it. By taking one step forwards,
you can eliminate the Coke can from the frame. We haven't changed reality,
or manipulated it afterwards, we've just chosen a slightly different framing
to show a slightly different bit of reality. Even the most extreme purists
would probably consider that acceptable.
This time, we reach
down, pick up the can and move it two metres away, where it is out of
the frame. We have altered reality. Is this different from changing our
own position? Is it acceptable?
Ok, now we reach down,
pick up the can and take it away with us, to throw into a litter bin later.
We have permanently altered reality. Is this different from moving it
out of shot? Is it acceptable?
Finally, we leave
the can where it is, and we use the healing tool to remove it from the
image. Is this different from picking it up at the time? If so, why?
We can ask the same
question with a zit on someone's face. It wasn't there last week, it won't
be there next week, is there any reason it has to be in the shot because
we took the photo this week?
Clearly there are
particular specialisms where the answers will differ. In forensic photography,
for example, we want to see things exactly as they were: nothing added,
nothing omitted, nothing distorted.
Photojournalism is
a specialism where we have a similar concern for accurate depiction of
what occured, but where things get more difficult. No news story describes
everything that went on - only part of it. When a plane crashes
on a town, we don't see the other side of town that wasn't touched, we
see the path of destruction. A photographer probably won't photograph
the uninjured people walking away from the plane with little assistance,
but is likely to photograph the ones covered in blood. It's hard to avoid
this kind of selectivity. But whether the manipulation occurs in the choice
of subject at the time, or in Photoshop afterwards, I think we can at
least agree that what is shown must be an accurate (if selective) record
of what was present at the scene.
At the opposite end
of the spectrum is an image which is pure art. Its only function is to
be aesthetically pleasing or thought-provoking. Whether the image has
any objective existence in reality, or was simply imagined by the artist,
is of no consequence.
It's between the two
that things grow more complex. A portrait is probably the most controversial
subject matter. Are we trying to portray the person as they are, or are
we trying to portray one aspect of them?
I'd argue that it
is impossible for a portrait to do anything other than depict a certain
slice of a person. Perhaps a better portrait gives a more detailed and
complex picture of someone than a corporate headshot, but the whole person
in one photo ... ?
In a portrait in particular,
the idea of objective reality gets pretty tenuous, in my view. Imagine
two people looking at the same person, let's call her Susan. One of them
is a stranger, the other is Susan's lover. Do they both see the same thing?
Clearly not. Now introduce a third person, who does not know Susan personally
but is a huge admirer of her work. That person sees something different
again.
So personally, I have
no difficulty with the concept that a photograph can only ever show one
slice of reality.
I guess I have a slight
prejudice towards effects which can also be achieved in a traditional
wet darkroom. I readily admit there is no rationale behind this, it's
simply based on familiarity. I do, though, love the fact that what used
to take an hour in a smelly darkened room can now be done in five minutes
in the comfort of my living room!
Here,
then, are the most common forms of post-processing that I do
to my images. The instructions in italics are for Photoshop CS, but should
be largely similar for Photoshop Elements.
Crop
Although you can crop
images as you please, just as we could in the darkroom, it is far more
satisfying to frame the shot correctly when you take the photo. Additionally,
every time you crop, you
are throwing away resolution. You paid for all those pixels, so it doesn't
make sense to discard them!
However, there will be times when cropping is unavoidable. For example, if you cannot get close enough to frame the way you would like to, or because you're shooting in a 3:2 aspect ratio and want to 10x8" print.
Photoshop: There are two ways of cropping in Photoshop - the crop tool, and the marquee tool (dotted-line rectangle). The crop tool is more flexible, allowing you to adjust to any shape rectangle, but the marquee tool can be handy as it allows you to specify (in the drop-down menu at the top) a fixed ratio. This is handy when the final output will be print.
Levels
Levels are a sophisticated
way of controlling both the brightness and contrast of an image. You have
a lot more control if you do this via Levels, but it's really no more
difficult than using simple brightness and contrast controls.
I tend to give Photoshop's
Auto-levels a chance, but usually end up doing it manually. I use the
histogram as a guide, but mainly judge it by eye.
Photoshop: To
try Auto-levels, press Shift-CTRL-L. If that doesn't look right, hit CTRL-Z
to undo it, then press CTRL-L to bring up the levels window. You have
three markers. The left one is the 'white-point', the centre one is the
'mid-point' and the right one the 'black-point'. If you drag the white-point
to the right, you will darken the image. If you drag the black-point to
the left, you will lighten the image. If you move the mid-point, you will
alter the contrast - right for more contrast, left for less. Use the histogram
as a guide to where the white and black points should be, but use them
as guidance not instruction. In particular, if you have (and want) solid
black in the photo, you will usually want to leave the black-point where
it is. You can also adjust levels by using the dropper tools inside the Levels box. Click on the black dropper and click something in the photo that should be solid black, and the white dropper and click something that should be solid white. You can also use the grey dropper to select a midtone.
Conversion
to black & white
Although a few digital
cameras have a black-and-white mode, the vast majority store colour images
which then have to be converted to black-and-white if desired. Even if
your camera has a black-and-white mode, I recommend shooting in colour
as you can then control the tone of the final image - a bit like being
able to choose the film you use.
There are three ways
of converting a colour image to black-and-white:
Greyscale
conversion:
This is the quickest and simplest, but gives the crudest conversion and
zero control over the tone. This is also what you get if you use an in-camera
b&w setting. It tends to give a high-contrast image which can look
good for architecture but doesn't look great for portraiture.
Photoshop: Select
Image » Mode » Greyscale
Desaturation:
This gives a wider range of greys, and is also very quick.
Photoshop: If
opening a RAW image, you have the option to drag the Saturation slider
all the way to the left (or enter a 0 in the box) as you open the image.
Otherwise, select
Image » Adjustments » Desaturate.
Channel mixing:
This is the most flexible conversion method, and the one I always use
for portraiture, and use most often for everything else. Although it will
take a couple of minutes to explain, it is worth the effort as you'll
get much better results. For portraiture in particular, the channel mixer
conversion method is nothing short of miraculous!
So, let's explain.
Digital images are captured as RGB images - that is, a mixture of Red,
Green and Blue dots. All other colours are formed by mixing different
levels of red, green and blue within a pixel. When we talk about the data
stored in the three primary colours, the phrase used is colour 'channels'.
Why do we care about
colour channels when we want a b&w image? Because the sensor in a
digital camera reacts differently to different colours. Sensors are most
sensitive to greens, so the green channel has the highest level of detail.
The least sensitivity is in the red channel, so we will have less detail
in this channel. The blue channel tends to have the most distortion, or
'noise'.
What channel mixing
allows us to do is select how much data is taken from each colour channel
when we create the b&w image. We can get any effect we want by mixing
the red and green channels, taking nothing from the blue channel ...
If we took 100% of
the data from the green channel, and nothing from the red and blue, we'd
have an image in which the detail shows up really well (better than in
the colour image).
If we took 100% from
the red channel, we would have an image that was quite soft, and lacking
in detail.
So, for an architectural
shot, we want a very high percentage of data from the green channel, and
not much from the red. For a portrait of a young woman, in contrast, where
we want to emphasise the softness of the skin, we would want a very high
percentage from the red channel.
A high proportion
of the red channel in a portrait tends to obscure any skin blemishes (because
marks & lines are detail that show up mostly in the green channel),
and also gives a kind of glow to the skin. It is a very easy way to create
a flattering b&w portrait.
All channel mixes
must add up to 100%. As a starting point from which to experiment, I recommend
the following mixes:
Architectural shot: 80% green - 20% red - 0% blue
Portrait of a man: 70% green - 30% red - 0% blue
Portrait of a woman: 30% green - 70% red - 0% blue
(I don't shoot landscapes, so can't help there, I'm afraid. I'd welcome
feedback from those who do.)
Photoshop: Select
Image » Adjustments » Channel
mixer. Tick the Monochrome box. Then adjust the percentages using
either the sliders, or simply typing the percentages into the boxes. Make
sure the percentages add up to 100%.
Healing
tool
The healing tool is
a very powerful effect that enables you to copy one part of an image and
paste it over the top of a different part. There are two main uses for
this: removing distractions, and covering up blemishes.
The classic example
is is a rugged landscape with a telephone wire running across one corner
of the sky. The healing tool allows you to copy a section of empty sky
and paste it over the top of the wire. Because you are copying texture
as well as colour, the illusion can be very convincing. If you are left
with any sharp edges, the blur tool will usually take care of those.
For portraits, you
may wish to remove obvious skin blemishes. Again, you paint over the blemish
with a section of smooth skin.
Photoshop: The
healing tool is the plaster icon on the toolbar (4th one down on the left).
Click it. Let's remove a spot. Alt-click (hold down the Alt key while
you click) on a patch of clear skin next to the spot. That copies that
part of the image. Check the toolbar at the top and make sure the mode
is set to Replace (not Normal). Then click over the spot. You may need
to adjust the size of the 'brush' in the toolbar at the top. Always Alt-click
very close to the bit you are covering, as changes in light and texture
will very quickly show up otherwise. You can 'paint' with the healing
tool, but you tend to have to do it in small sections, Alt-clicking to
select a nearby surface each time.
USM
(Unsharp Mask)
Images
from digital cameras tend to be a little 'soft' (very slightly blurred
edges). Sharpening is the process of reducing this softness. Don't worry
about why a sharpening technique is called Unsharp Mask - it's just a
historical quirk from the method used in darkrooms.
Like colour saturation,
people can get very carried away with sharpening. Some images are so ridiculously
over-sharpened that they look totally artificial, as if someone has cut
out the subject with a pair of scissors and glued it onto the background.
It is far, far better not to sharpen at all than to over-sharpen.
Photoshop: Select
Filters
» Sharpen » Unsharp mask.
There are numerous tutorials on the web that will explain the meaning
of the three values, but for this beginner's guide, just try 85-1-2 and
see how that looks.
Gaussian
blur
It may seem an odd
concept to spend large sums of money on sharp lenses and then sometimes
choose to introduce an artificial blur, but sometimes that's what we want
...
The most common use
of gaussian blur is for portraits of women. A DSLR like the
D200 is perfectly capable of showing every pore in the skin. Lens geeks
may like this, but portrait subjects - especially women - tend not to.
;) Gaussian blur is an effective way of making the skin look smoother,
reducing both blemishes and lines. (For b&w photos, you may well find
that this isn't necessary - the channel-mixer approach described above
is all that is needed.)
I typically use a
radius of 2 and an opacity of anything from 40% to 100%. I then use the
history brush to restore the eyes, lips, hair and edges of the skin (as
well as anything else in the shot).
Photoshop: You must do this before you make any other alterations in Photoshop (as otherwise the History brush will be unavailable). Select
Layers
» Duplicate layer and name it Gaussian blur. Select
Filters
» Blur » Gaussian.
Set the radius to 2. You will see that the image now looks horribly
blurred - don't panic! Click on Opacity and slide the slider left and
right to see the effect. Typically, I'll select 40%, but sometimes go
as high as 100%. The photo will still look horribly blurred, but that's
ok as we're going to remove the blur from everywhere except the skin.
Click Ok. Select the History brush (5th icon down on the right). Select
a fairly small brush size. Now brush over the eyes (including eyelashes
and eyebrows), lips (and teeth, if showing), hair and all the edges of
the skin (be careful to do just the edges - use a small brush for this).
Select a larger brush to paint over the background to restore the natural
bokeh (blurredness) rather than the artificial one we created. What you
are doing is removing the Gaussian blur layer from the areas you paint.
So ... that's my brief
introduction to post-processing. Hope it helps! In terms of developing
your skills from here, there are shedloads of Photoshop tutorials on the
web.
|
