Optical density to grayscale values

Hi,

I am trying to use a VIDAR VXR Film Digitizer and 32-bit TWAIN for
windows to make a .tif file which is an absolute faithful conversion
of all the optical densities on a radiographic film into grayscale
values.

The reason is that I want to make distance measurements on my .tif
file from a point of particular optical density to another point of
optical density - not from one physical structure to another.

I know that the software can do clever things to improve contrast or
brighten a dark area but in this case I just want my .tif file to look
exactly like the radiographic film with no enhancements.

What settings in the TWAIN software and the scanner (digitizer) should
I use to achieve this.

For example, I have the option of using Translation tables:
Histogram-equalized
Linear histogram-based
Linear histogram independent
Linear optical density
etc.

Other options include 'dark enhance' and resolution.

Any advice is greatly appreciated,

Chris.

"Chris" <c_g_taylor2000@yahoo.co.uk> wrote in message
news:33598cff.0409070824.518d8fde@posting.google.com...
> Hi,
>
> I am trying to use a VIDAR VXR Film Digitizer and 32-bit TWAIN for
> windows to make a .tif file which is an absolute faithful conversion
> of all the optical densities on a radiographic film into grayscale
> values.
>
> The reason is that I want to make distance measurements on my .tif
> file from a point of particular optical density to another point of
> optical density - not from one physical structure to another.
>
> I know that the software can do clever things to improve contrast or
> brighten a dark area but in this case I just want my .tif file to look
> exactly like the radiographic film with no enhancements.
>
> What settings in the TWAIN software and the scanner (digitizer) should
> I use to achieve this.
>
> For example, I have the option of using Translation tables:
> Histogram-equalized
> Linear histogram-based
> Linear histogram independent
> Linear optical density
> etc.
>
> Other options include 'dark enhance' and resolution.
>
> Any advice is greatly appreciated,
>
> Chris.

There is no consumer level scanner that can be relied on to make optical
density readings reliably. The line sensor is not linear.

The best that you can do is adjust the Black point and White point then scan
and adjust in photoshop.
Then all densities will fall within the range of the scanner. But they will
not be the optical density of the original.

If you have or get a Kodak gray scale strip and scan it, you can get a good
idea of the scanner's response to a given density.

These are the best images I could get with my flatbed scanner of a Kodak
color patch and gray scale and a MacBeth ColorChecker.

Both images are Tiff when you download. The Tiff images will not show in
Internet Explorer.
http://www.carlmcmillan.com/Test_images.htm

--
CSM1
http://www.carlmcmillan.com
--

"Chris" <c_g_taylor2000@yahoo.co.uk> wrote in message
news:33598cff.0409070824.518d8fde@posting.google.com...
> Hi,
>
> I am trying to use a VIDAR VXR Film Digitizer and 32-bit TWAIN for
> windows to make a .tif file which is an absolute faithful conversion
> of all the optical densities on a radiographic film into grayscale
> values.
>
> The reason is that I want to make distance measurements on my .tif
> file from a point of particular optical density to another point of
> optical density - not from one physical structure to another.

For a radiometric application, you need to start with a linear LUT.
That LUT needs to be calibrated for the type of film you evaluate
(single/double emulsion, film speed). The linear film transmission
data can then be converted to relative optical density. From the
descriptions given, a "Linear optical density" seems to do that in one
LUT operation.

> Other options include 'dark enhance' and resolution.

Don't use dark enhance (it will destroy the Linear/Log data
relationship), and select the resolution that provides the best
Signal/Noise ratio (presumably the lower resolutions), because you
could be confronted with high film densities.

Bart

CSM1 wrote:
>
> Both images are Tiff when you download. The Tiff images will not show in
> Internet Explorer.
> http://www.carlmcmillan.com/Test_images.htm
>

The tiffs open up fine if you use the AlternaTiff plugin.
http://www.alternatiff.com/

Eric

> For a radiometric application, you need to start with a linear LUT.
> That LUT needs to be calibrated for the type of film you evaluate
> (single/double emulsion, film speed). The linear film transmission
> data can then be converted to relative optical density. From the
> descriptions given, a "Linear optical density" seems to do that in one
> LUT operation.
>
> > Other options include 'dark enhance' and resolution.
>
> Don't use dark enhance (it will destroy the Linear/Log data
> relationship), and select the resolution that provides the best
> Signal/Noise ratio (presumably the lower resolutions), because you
> could be confronted with high film densities.
>
> Bart

Ok thanks, what is a linear LUT, and what do you mean by "the linear
film transmission data can then be converted to relative optical
density".

I have a default Linear Optical density translation table.

I also tried making my own translation table - a text file with a
single number on each line starting at zero and then down to 4096.

"Chris" <c_g_taylor2000@yahoo.co.uk> wrote in message
news:33598cff.0409090504.1a59991@posting.google.com...
SNIP
> Ok thanks, what is a linear LUT,

Linear Look-up Table. Instead of calculating output values for each
input value, it is quite common to use a pre-calculated table for all
possible input values.

> and what do you mean by "the linear film transmission data
> can then be converted to relative optical density".

The film transmission is recorded by the CCD as an linear transition
in charge. Twice the amount of transmission is twice the amount of
charge. That charge/voltage is converted/quantized by the Analog to
Digital Converter (ADC) to (presumably) a linear discrete output
value. Any non-linear response, either caused by the ADC circuitry
and/or the type of illumination and interaction with grainstructure of
the film (=Callier effect), must be calibrated to a known linear
response.
Optical density is calculated as the Log base 10 of the reciprocal of
Transmission, but it assumes the transmission values are calibrated to
real transmission differences.

> I have a default Linear Optical density translation table.

If the software was used to calibrate the system to the film, then the
prior OD calculation is already done, and the result was preloaded in
a LUT (the Linear Optical Density translation table). But you'll have
to check if that is exactly the way it was implemented with that
scanner. If you check with a stepwedge (appropriate material, KV and
A) exposed on the x-ray film you can verify if it all works as
expected.

Bart