|
If you are primarily interested in collecting album shots then
processing may very well be unnecessary. Just print your JPEG images the way they
come out of the camera or have a photo-shop do it from your memory card. Camera
processing is usually very good. If you want to exhibit your work in
larger form, where it will be subject to critical examination and appreciation,
you will need to get involved with the "digital darkroom". Some
software suggestions have been made elsewhere.
If you shoot JPEGS or TIFFs, in-camera processing has already been applied
and you are, furthermore, stuck with 8 bits/channel ("24 bit") images. These may be very
fine images indeed but they are usually a compromise you could improve upon. The
in-camera processed images can always be further touched up but the potential is
somewhat limited as the processing already applied may not be reversible. For your best work, shoot RAW format (in Nikonese, NEF)
at the highest available bit-depth and do
your own processing. RAW/NEF bypasses all camera processing with rare exceptions
(Sony Alpha, for example, which applies some noise reduction).
Another big advantage of doing your own processing is that you can, at last,
in small format photography apply the "Zone System" of exposure and development
as first described by Ansel Adams for optimum rendition of large sheet film images.
This technique requires individual processing for each image -
impractical to achieve with 35mm film or other strip media.
Storage is cheap. Never, ever, risk your originals, just as they come out of
the camera. Once an image has been processed in any way it is usually difficult
if not impossible
to reverse what has been done after you have SAVEd it. Make a backup or adopt some
naming standard that keeps originals untouched. DVDs are practically free and
online mass storage such as Western Digital's "My Book" is inexpensive. You cannot turn
back the clock and recreate the exact circumstances of
your best picture taking moments. Furthermore, software is always improving and
the day will almost certainly come when you want to start from scratch with some
of the latest processes applied to your favorite images.
Silkypix is very civilized in that it absolutely refuses letting you write
over a source file. With other products you may have to be careful. Picture
Window's "pipeline" editor automatically generates new names at each processing
step so when you save, overwriting the original requires an explicit act of
destruction. These are responsible, professional safeguards.
Do the least amount of processing that meets esthetic objectives and the
requirements of presentation (computer monitor or printer). Too much
sharpening, noise reduction, histogram expansion and so forth invariably leave tell-tale
tracks in the form of halos, posterization and other distracting, amateurish defects.
If you have a sequence of steps that produces the results you seek, package
it up as a script or profile to save for later use. This largely eliminates the
tedium of respecifying processing steps time and again. You will probably end up
with a library of profiles so put some effort into a good naming standard (for
example "Cloudy_Days_Scenery.spd, Incandescent_Lighting.spd, etc.). Most
processing software has batch and profiling capabilities so you can run a profile against any number
of images.
Learn more about this in the "Color Management" section.
If an image has low noise to begin with, bypass this step. It is better to
leave a small amount of noise than trying to remove all of it and smudging fine
detail. The foundation of realism in an image is detail and you want to keep as
much as possible. DSLRs have inherently low noise so it's best usually to bypass
this step altogether if that's the type of camera you have.
Unfortunately, some software capable of working directly with RAW has weak
noise reduction capabilities. Picture Window comes to
mind. The solution is to convert the image to an unprocessed 48 bit
TIFF, run Noise Ninja or Neat Image and then finish processing the image in TIFF
format - assuming you need noise reduction at all. This is a minor nuisance but an effective solution. It's really best
just to use a low noise camera to begin with or low ISO settings on your
otherwise noisy
camera.
You will find some variance concerning what to do next depending on who you
read. The following order is a suggestion only and minimizes interactions.
Silkypix presents its processing tools in this order as well (apart
from deferring noise reduction until later). Your
images might already be nearly perfect. Skip the actions your eyeball informs
you are not needed:
- Exposure bias and highlight control (these interact and also
interact somewhat
with contrast) - The live histogram view in your imaging software is a big
help with this because you can easily see if shadows are being blocked up,
highlights at risk of being blown, etc. Silkypix has a superb highlight
controller that helps you recover detail from very bright parts of an image
and restore other essential properties.
- Color balance ( Paintshop and Silkypix have excellent color-cast removal) - Outdoor shots around noon will have a bluish cast you will probably
want to remove, incandescent shots will be too red, etc.
- Contrast and Gamma - The histogram may show
that the image tonal scale is too short and you might want to stretch it
out. The risk is posterization if you overdo it. Watch the histogram. If it
starts showing prominent, evenly spaced peaks you probably have gone past
what the bit-depth of the image can support. Gamma is a refinement of simple
contrast. It affects primarily the rate of change of mid-tones. These settings may prove critical to print
appearance. Experimentation
will suggest how much you should compress/expand the tonal scale (if at all).
- Saturation - If you decide to increase color
saturation, watch the histogram to ensure you do not clip any of the color
channels.
- Sharpness - There are a number of methods
available for recovering detail. Silkypix and Paintshop do an exceptionally good job. "Unsharp
Mask" (USM) gets a lot of attention but over-use results in ugly halos
around high contrast areas. I prefer "hi-pass" sharpening for most images.
Silkypix doesn't make USM available until time comes to save an image, then
pops up a secondary preview window. This is wise because it discourages
overuse of USM.
- Clone brush - to remove any specks or other
gremlins.
- Other "mechanical" adjustments
- Perspective
correction, barrel/pincushion distortion etc.
Sharpness is left until later because the other processes affect
fundamental detail in unpredictable ways. Clone brush is last because sharpening
and contrast have a way of causing specks to magically re-appear if removed
early-on.
Some iteration is inevitable as most steps will interact a bit. Sharpness
usually requires a touch-up prior to printing and after re-sizing (if you choose
to do this yourself rather than leaving it to the printer). It's all something
of an art and you will do better at processing with lots of practice.
The Internet presentation standard for color is sRGB so make sure any
images you want to share conform to this color space specification.
There's considerable variability in the quality of images one sees on
Internet photo sites. On one hand, you want to keep size small for fast
download but on the other, small file sizes can mean loss of detail and
poor color. I find 8 x 10 images an excellent compromise among viewing
comfort, download convenience and presentation characteristics although
smaller sizes work well too, depending upon theme. The preparation steps (assuming the image
processing has been completed) are as follows, using my imaging
software. The steps must be taken in the order shown:
- Resize the image so the long side is 10 inches (or whatever you think
looks best - 4 to 8 inches is good for many pictures). Simultaneously re-sample to 100 dpi with bi-cubic interpolation (or
whichever interpolation method looks best to you). Bi-cubic minimizes
staircase effect and in my experience produces the most visually pleasing
re-sampling for images to be viewed on a monitor.
- Optionally place narrow borders around the image (it looks better that
way but it's up to you). Black and/or white provide excellent visual control
points for helping the eye judge tone and color. Borderless images can be
effective, provided they are placed on a neutral gray or black background.
Personal taste comes into this. Colored framing can work - provided you have
"the eye" for picking something that complements the image. I don't have
this talent so stick to colorless borders and backgrounds.
- Convert to JPEG with 20% compression. This will yield images generally
not exceeding 400 mb for a 10 inch long side and a good deal less for
anything smaller. The JPEG losses will be invisible.
You are now ready to upload. In case
you are wondering, compressing before re-sizing will almost certainly
result in objectionable contouring of tones. That's because to get from
the several megabytes original to an acceptable download size (500k or less) you
have to use compression in the 90% range. It looks terrible.
Why 100 dpi? The Internet
"standard" is 72 dpi but this is an anachronism. Most fixed-pixel
monitors today (LCD's for the most part) have almost exactly 100
pixels/inch. To take advantage of this for image viewing, the following
conditions must be met:
- The viewer's monitor should be set to native resolution. That is,
if the monitor is rated at 1920 X 1200, then any setting less than
that could result in the image being enlarged upon presentation with
each image pixel being represented by more than one monitor pixel
(re-sampling will be used to get a proper fit). Drop the monitor's
resolution by too much (1024 x 768, for example) and image
deterioration will be pronounced.
- The browser zoom settings play a role. If other than 100%, more
re-sampling with potential for degradation takes place. Zooming out
to re-size an image that is too large usually works well.
It is impossible to know what your site visitors' settings will be so
for the best viewing experience you should probably preface your work
with recommended settings or, at least, a note concerning what to expect
if the monitor is not in native mode.
There will always be someone who finds your choice of scale and
resolution unsatisfactory so do not obsess over a perfect solution. If
you really want to accommodate everybody, consider offering images at
differing sizes and resolutions, selectable by means of an on-screen
option. This is a lot of work and you might want to think twice about
going that route. Another way is simply to produce small images (5 or 6
inches maximum on the long side). These should present no difficulty to
anyone but will not exploit the full capabilities of high performance
monitors.
Point out to visitors that browsers have a zoom control and this is
often the most satisfactory way to scale inconvenient images.
In Paintshop I have scripts that fully automate re-sampling and scaling. I am usually
working from full resolution TIFFs prepared from NEFs. At this size and
resolution (8 X 10, 100 dpi, for example) a JPEG compression of 20% (this yields a very high
quality output) gives a file sizes ranging from 150K to 400k which is quite
manageable for most visitors. If I want something under 150k for the
bandwidth-challenged or if uploading to sites with file size restrictions, 50%
to 60% can still look half-decent if just a bit soft. The reason for explicitly
specifying a real-world scale (e.g., 8 X 10) is that this helps viewing software
that doesn't know how to scale images for a monitor. There's still some like
that around. These images will look just fine on any monitor without additional scaling. On
lo-res monitors images may appear a bit too large so if you want to cater for
those situations, reduce the physical dimensions accordingly. There's no
one-size-fits-all solution so some compromise is unavoidable.
These pictures will have all the detail a monitor can render but will not
withstand any magnification/zooming.
How good do these images look?
Click here.
Back to top
of page
|