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

On Mon, 23 Jun 2008, Tina Carvalho wrote:

> Back in the Really Old Days the development of negatives and the
> production of photographic prints were detailed in the Materials and
> Methods of a paper, right down to the concentration of developer, minutes,
> etc., plus any dodging and burning. This eventually got shortened to just
> negative and paper type and grade, and shortened more, until not much was
> reported. Made for easier reading. I don't miss those days. BUT if we now
> go back to more reporting of the details of image acquisition, it will
> serve to promote reproducibility and, one hopes, make people think about
> what they're doing so that they do not unintentially manipulate their
> data.
>

The big problem, frankly, is that it is fundamentally impossible to create a digital image without image processing.  The creation of the image itself requires exactly the kind of processing that some folk call fraud.  The only difference is *where* you do the processing -- in the camera, in the computer, or in the display/printer.  The implication is that image processing is OK as long as it's done ignorantly.

The idea that there is some sort of pristine image in the camera sensor that must be preserved is analogous to claiming that it is fraud to develop and print photographic film because both the processing and the printing require "processing" the image -- dealing with issues of contrast, brightness, color balance, etc., etc., etc..  To claim that analogous processes in digital imagery is some sort of fraud is silly.  Is color calibration "fraud?"  I think not.  In fact, I think a better argument can be made that there could be a greater misrepresentation without color calibration, gamma correction, etc.

Attempts to say you can't do such things is merely requiring that the user be willfully ignorant of what's happening and use the default parameters (or last set parameters) of whatever imaging pipeline is used.

As I've noted before, this was a hot topic in the forensic imaging world a few years ago.  One set of best practices guidelines adopted by many forensics labs (the Scientific Working Group on Imaging Technologies, or SWGIT) guidelines are pretty explicit.  See:

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

and look at Section 11: Best Practices for Documenting Image Enhancement

Basically the document divides enhancement methods into basic processes commonly used in image production and/or analogous to commonly accepted darkroom film processing (brightness/contrast adjustment, cropping, rotation, inversion, white balance, file format conversion, etc.) and "advanced" techniques such as deblurring, noise reduction, image restoration, etc.

For basic processes, it is only necessary to indicate that they are done through some SOP or similar documentation.  For advanced processing, all parts of the pipeline must be documented:

"Documenting image enhancement steps should be sufficient to permit a comparably trained person to understand the steps taken, the techniques used, and to extract comparable information from the image. Documenting every change in every pixel value is discouraged because it adds nothing of value to the analysis.

Exploratory enhancement operations not incorporated in the final image do not need to be documented. Test prints and/or intermediate images resulting from a variety of techniques not incorporated into the final image should be discarded.

Minimum requirements for documentation include identifying the software application and/or techniques as well as the settings and parameters used. Automated processes, such as running user-defined macros, require only documenting usage if the process is
defined in the agency documentation."

billo