Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal

On Sat, 16 Jun 2007, Glen H MacDonald wrote:

> Search the CONFOCAL archive at
> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>
> Dear Jerry,
> This can of worms squiggles most strongly exactly do to the reactive nature 
> of the perception of image processing.  Photoshop is not the problem.  Not 
> the solution, either.
>
> Image processing should be approached as an act of translating the signal 
> detected and produce by an instrument that responds to intensities into a 
> representation that can be recognized by the human visual system designed to 
> detect edges, changes and contrast.  Image process is all too often viewed as 
> random actions to address personal esthetics. The entire process of recording 
> a microscopic image is a series of translations, beginning with translating 
> the sample into a diffraction pattern at the back of the objective lens, then 
> into electrons, into a series of digital values.  We spend huge amounts of 
> money, time and procedures to ensure that these translations are carried out 
> with minimal losses at each stage of the instrumentation.  Then most people 
> screw it up at the very end by using arbitrary actions in image processing 
> software to make something look "good" (meeting pre-conceived notions).
>
> Information losses and histogram changes occur at every stage, most affect 
> gamma.  Even the choice of excitation/emission parameters (filter, laser 
> lines, etc) will skew the gamma.  As you point out, the resulting image must 
> be processed in some manner to both compensate for instrument effects as to 
> render it meaningful to our visual system.
>


A similar discussion has been a serious issue in forensic image
processing, both in terms of ethics and in terms of admissability.
In the late 90s the FBI assembled a group representing a broad
range of players including criminalists, crime scene technicians,
forensic and intelligence image analysts, academicians, and others
from the US with input from agencies in the UK, Canada, Australia,
the Netherlands, and other countries.  The group, called the
Scientific Working Group on Imaging Technologies, developed some
basic best practices for image processing in the forensic arena.

The conclusion we came up with was that the issue was not whether
or not image processing was done, but that the image processing 
was adequately documented.  We defined "adequate" as being documented
to the degree that a similarly trained professional could read
the report and obtain a similar conclusion (e.g. not a bit-for-bit
reproduction, but a diagnostically similar result).  We classified
image processing methods into "basic" and "advanced."

Basic processing was defined as those processes that have trivial
correlates with chemical processing/printing and that any professional
should comprehend easily.  It is sufficient to simply say that
these processes are performed without having to document parameters,
specific steps, etc.  For instance, one can simply say that "color
balancing was performed."  However, if one does it, one must say it.

Advanced processes, such as deconvolution, removal of periodic
distractors by manipulating the power spectrum, etc. require
more complete documentation, but again, not the keystroke-level
logging desired by some trial attorneys.

The SWGIT has promulgated a number of "best practices" documents,
some of which have become the basis for accredation requirements
by the ASCLD-LAB (American Society of Crime Laboratory Directors
Laboratory Accredation Board).  See:

http://www.theiai.org/guidelines/swgit/index.php

In particular, you might take a look at Section 5 (Recommendations and
Guidelines for the Use of Digital Image Processing in the Criminal
Justice System ) and Section 11 (Best Practices for Documenting Image
Enhancement).

billo