Digital Work Flow
by Roger N. Clark
All images, text and data on this site are copyrighted.
They may not be used except by written permission from Roger N. Clark.
All rights reserved.
In the new digital imaging era, there are many ways to process an
image. On this page, I'll discuss the work flow I generally use in
processing my images. I must stress that these are my opinions.
There are often multiple "correct" ways to process digital image
data. and many can produce excellent results.
Digital work flow starts at the point the data are converted
from analog to digital. With film this is the step starting at the digital scan.
With a digital camera, it is the settings of the digital
camera that begin the work flow.
Film. I scan all my film at the full precision of the scanner.
Good scanners are at least 12-bits/channel. The output file is 16-bits
per channel TIFF images. I do the minimum processing at the time of
the scan and save corrections for the photo editor where I have
more control. Scan parameters include:
- Straight-line transfer curve.
- Brightness correction only.
- Little to no color correction. (I try for correct color at the
time the film is exposed, using color correction filters, if
necessary.)
Digital Cameras. I do both jpeg and raw format output. Jpeg is
only 8-bit, while raw on many cameras is 12 bit and a few now have 14-bit
output. Raw files are converted to 16-bits/channel TIFF files.
Critical to the digital work flow is backing up the digital files.
- I try and carry enough compact flash memory cards to cover at least
one day of photography.
- I carry two portable hard drives and back up each day's images
to both hard drives. These are independent systems with card
readers and screens for viewing images. Sometimes I carry a
laptop and USB drives instead. The independent systems also
act as USB drives for later transfer of data to a computer.
- I do not delete the contents of a compact flash card until I have
filled all cards and need more memory. Then I delete each card
as needed. (Formatting a card clears the card faster than deleting all
the images.)
Archive the Original Image Files
It is very important to archive all original image files without any
modification, whether they be film scans, jpegs from a digital camera,
or raw files from a digital camera. You will see below that I save
copies of the original in directories I call "orig." I never modify
the original image files. I only modify copies of those files, and I
never use the same name in order to prevent any version from copying
over a different version. The reason for archiving the original images
is that in case you want to go back and redo the processing, you can go
back to the original file. Maybe your expertise will improve, or maybe
a new processing algorithm with be invented that makes the image better.
You need that original file to take advantage of these possibilities.
Digital Image File Management
People often ask about image management software. I do not use any.
One reason is that I do not want to be locked into a proprietary system
that may not be around in a few years. I want to be able to find my
images in 1 year, 2 years, 10 years, 30 years, and for my descendants
to find them too, without special software.
I could use an emerging technology of embedding keywords into the image
itself (called tags) and then using programs that build ascii lists of file names and
key words. That would work and be portable and readable far into the
future, except for one problem. Some programs I use do not propagate
the embedded tags, so the tags would get lost. There is no single software package
that does all tasks. For example, photoshop does not do sophisticated
sharpening (e.g. Adaptive Richardson-Lucy deconvolution), and software that
does ignores some metadata.
So I organize my photos according to subject by using long descriptive
file names and using a directory structure that isolates subjects.
For example, I may do something like the following.
- colorado.2005/georgetown/bighorn.sheep.c12.25.2003.img_7089.b.tif
- colorado.2005/georgetown/orig/bighorn.sheep.c12.25.2003.img_7089.a.crw
- family.2005/christmas/family.portrait.c12.26.2005.img_1234.c.tif
- family.2005/christmas/orig/family.portrait.c12.26.2005.img_1234.a.crw
- alaska.2004/katmai/bears/brown_bear.c09.07.2004.JZ3F0862.b.tif
- alaska.2004/katmai/bears/brown_bear.c09.07.2004.JZ3F0862.b-700.jpg
- alaska.2004/katmai/bears/brown_bear.c09.07.2004.JZ3F0862.b-16x24LJs.tif
- alaska.2004/katmai/bears/orig/*.crw
- alaska.2004/kenai/eagles/eagle.c09.10.2004.JZ3F4428.b.tif
- alaska.2004/kenai/eagles/orig/eagle.c09.10.2004.JZ3F4428.a.crw
where the "/" means directory boundary, so colorado.2005 is a directory.
The image catalog number, e.g. c12.25.2005 is the month, day and year,
and the 4-digit number is the camera's image number. The letter
code before the file extension is the processing version. The brown bear
image above with b.tif is the final processed image, version b, the
b-700.jpg file name means version b with 700 pixels, signifying
my web image. The brown bear image with b-16x24LJs.tif is version b scaled to
a 16x24 inch print file for a Lightjet Fuji Crystal Archive ICC profile with signature.
The "orig" directory holds the original untouched files, whether jpeg, tif, or
raw (e.g. .crw).
The above scheme allows me to quickly find any of my tens of thousands
of images. Between directory names and subject file names, I can
use the operating system's search tools to search key words and
find what I need among all images.
In January 2006, I have over 144,000 images online on my system.
In December 2015 I had 582464 images (it takes my several year old 3.07 GHz I7-950 computer
less than 1/4 second to find images).
With a few line linux/unix script, I build a simple ascii listing of images
anywhere on the system. I can then search that database in a fraction of a second
with a one line unix grep command. Example
scripts are shown here.
These scripts work on linux, macs and can work on windows 10 if you install
the linux tools from microsoft: https://msdn.microsoft.com/en-us/commandline/wsl/about
which gives a linux bash shell and common linux commands.
In the photo editor
- 1) I only do 16-bit editing. If the starting file is 8-bit (e.g. jpeg),
the first step is to convert to 16-bits/channel.
- Why? with integer math, there is
always round-off error of 1 bit. First, the value for each
channel in a digital image file is called DN, for "data number."
For example, if you use the levels tool
to adjust the maximum level from DN 253 to 255, the software multiplies
the value of every pixel by 255/253 = 1.007905139. Because the result
is an integer, there is a plus or minus 1 digit round-off. In an 8-bit
file, this added error effectively means you have only 7 bits of
precision, or one part in 128. With a 16-bit file, you have 15
bits of precision, or one part in 32768. (Photoshop does 15-bit
math, so precision drops to 14 bits, or one part in 16384.)
If you do multiple editing steps,
added errors can result in poor intensity precision with an 8-bit
file. This is called posterization. 16-bit editing provides
enough precision so that posterization is not a
problem.
- Depending on the image, 8-bit versus 16-bit editing may not
be noticeable. The higher the signal-to-noise ratio, the more often,
in my experience, does 8-bit editing show posterization limits.
- 2) Color Space. Check the color space and convert to a wide color
space if not already there. I generally use Adobe RGB 1998.
- 3) Adjust Levels. First adjust the brightest portions
of the image to your liking using the levels adjustment tool. For example,
in a landscape image I might examine the brightest areas of the images,
like clouds and adjust the levels adjustment slider so that the brightest parts just reach
a value of 254 or 255 on the 8-bit scale typical of the slider tool.
(Photoshop only displays values 0 to 255 in the levels tool regardless
of the bit depth.) This levels adjustment is done on the entire image.
- 4) Curves Adjustment. The next step is curves adjustment using the
curves tool. Never use Photoshop's contrast and brightness adjustment tools
as they are additive. The curves tool, like the levels tool is a multiplicative
tool. Multiplicative has the correct math to mimic changes in scene brightness,
exposure, or f/stop changes. In the levels tool, the upper slider is used to
derive the multiplier, and lower slider is an offset (a subtractive adjustment),
and the middle slider changes the multiplier to a 2-part piecewise line
multiplier.
- 5) Dodge and Burn selected regions. Select different regions as desired
that may be at the limits of dynamic range and dodge and burn to bring them
into printable range. For example, bring up shadow detail, or darken clouds.
First select the area with one of the selection tools. Next feather it.
Feathering makes the selection area a smooth transition to the rest of the image.
Once feathered, adjust the level with one of the three following tools:
- Curves tool,
- Levels tool, or
- Shadow highlight tool.
- 6) Increase Image Size. If I'll be making larger prints, I'll interpolate
the image to a higher pixel count. I usually use bi-cubic or bicubic smoother
interpolation in Photoshop.
- 7) "Sharpen." The main tool many use is called unsharp mask, a filter.
But unsharp mask does not actually sharpen! The way unsharp mask works is that
the image is blurred using a certain radius, and then the original image is
differenced with the blurred image, and the result is added back to the original.
The effect is to modify contrast around edges. The amount added back is
usually controlled by the user (called amount in Photoshop). The effect is
actually a change in acutance, not sharpness. But increased acutance
gives the appearance of increased sharpness.
However, there are other methods that actually do sharpen. I use Richardson-Lucy
image restoration. Photoshop currently does not have this tool, so I write the
image out to a 16-bit/channel TIFF file and start another program to do the
job. For an example, see:
Image Restoration Using Adaptive Richardson-Lucy Iteration.
With a quality, low noise digital file, I have found I can increase my
print size by about 2 times (each dimension) using Richardson-Lucy
image restoration. For example, I find I can make nice sharp 16 x 24-inch
prints from 8-megapixel DSLR images.
- 8) Reassign Color Space, if necessary. If the image was sent to another
program for up-scaling, like in ImagesPlus in the step above, that does not do
color management, then the color space needs to be re-assigned. So, back to
Photoshop, re-assign color space.
- 9) Hand Editing. Hand edit grossest interpolation artifacts and fix other
defects, like dust spots, and film scratches (for scanned film).
- 10) Save the file. Depending on my intended use, I save as a 16-bit/channel
TIFF files, or convert to 8-bits/channel and save as jpeg. For best quality, 16-tiff
files are necessary. I only save as jpeg at highest quality and only when I do
not need highest quality.
- 11) Prepare the image for printing.
- a) Resize the image for a desired print size.
- b) "Sharpen" with unsharp mask.
- c) Convert to 8-bits/pixel.
- d) Convert to the color profile of the printer.
- e) Preview in proof colors, and check out of gamut.
- f) If proof colors or out of gamut is in error, fix.
(This is a major subject by itself and may be
treated in a future article.) Fixes include
curves adjustment, and conversion to LAB or CMYK
space and adjusting individual channels (often
of small areas, selected and feathered then fixed).
Note: out of gamut will not necessarily result
in a bad print. A good printer will simply
choose/use the nearest color it can print.
Here is an example of an image with a lot of
out of gamut (for a Fuji Crystal Archive Lightjet print:
Colorado Fall Serenity: Last Dollar Road, image L4.9362
This image produces a gorgeous 20x30+ inch print
on a Lightjet despite many of the oranges being out
of gamut.
- g) Save file (named differently from the above)..
- h> Make an inkjet test print.
- i) Resize file from step 10 for an inkjet test print
- ii) Convert to 8-bit.
- iii) Convert to the color profile of the printer.
- iv) Make print. If not OK, fix similar to step 11f.
- h) Write CD for lab print.
- 12) Prepare image for a web site.
- a) Open completed image 16-bit TIFF file.
- b) Resize to desired web size. I usually make the maximum
dimension 600 or 700 pixels.
- c) "Sharpen" with unsharp mask.
- d) Convert to 8-bits/pixel.
- e) Convert to sRGB color space.
- f) Add copyright as a new layer.
- g) Save as a Photoshop file (as a different name
from the original file.
- h) Save for web from Photoshop. This function reduces
unnecessary header information in the jpeg file.
I usually save with jpeg quality around 70%.
Backup the Image Data. I make multiple copies of the images I
want to archive. I write images to DVD and hard drives. I employ USB/Firewire
external hard drives and back up my images to the hard drive. I make a minimum
of 2 sets of backups on DVD and hard drives and keep one set off site. This
means I have 5 copies of the data: 2 DVDs, 2 offline hard drives, and on the
desktop computer. I also have converted my old PC to Linux servers with large
disks. I back up images to the servers for more copies, resulting in
3 different media/format types. This ensures against loss due to changing
technology and changing formats/file systems.
Notes:
DN is "Data Number." That is the number in the file for each
pixel.
16-bit signed integer: -32768 to +32767
16-bit unsigned integer: 0 to 65535
Photoshop uses signed integers, but the 16-bit tiff is
unsigned integer (correctly read by ImagesPlus).
Thus, Photoshop has 15-bit precision with 16-bit files.
http://www.clarkvision.com/articles/digitalworkflow
First Published 2005.
Last updated November 3, 2016