Digitizing X-radiographs

Rachel Danzing <rachel.danzing [at] brooklynmuseum__org> writes

>We are interested if anyone has digitized their x-radiographs and
>what their experiences were.  Were any equipment or special
>techniques found to be useful?

We have been scanning X-radiographs  of paintings here at the
National Gallery, London for a while now and the results have been
very useful. Most of the details of the process used here have been
written up in:

    J. Padfield, D. Saunders, J. Cupitt and R. Atkinson.
    Improvements in the Acquisition and Processing of X-ray Images
    of Paintings, National Gallery Technical Bulletin 23, pg 62-75,
    2002.

X-ray plates can hold a large amount of information, but they are
fragile, if they are digitized well people will generally never have
to handle the originals again, which makes keeping them intact much
easier.

X-ray plates have a very high dynamic range, (the number of gray
steps between black and white), the highest I have seen so far is
about 40,000 different levels of gray, which is a lot of
information.  Standard scanning will normally produce an 8-bit
image, i.e. an image with 2 to the power 8 levels of gray, which is
only 256 levels of gray.  Therefore, if you scan an X-ray plate at
8-bit you can lose an awful lot of information.  This problem can be
solved by making sure that your X-rays are scanned at 16-bit (2 to
the power 16 or 65536 levels of gray.)  Make sure that when they are
scanned that they are not scanned at 8-bit and then just saved at
16-bit, some scanner do this for some strange reason.

Resolution:  When you start scanning X-ray plates at high resolution
you can start to produce very big files, these need to be processed,
stored, archived and then manipulated for use.  If you begin to
mosaic groups of X-ray images together your files can get very big.
Before embarking on a large scale scanning project have some samples
done at different resolution, then compare then to see what level of
details is being captured.

The file sizes you mentioned in your post appear to relate to 8-bit
images.  As an example: painting 40 x 33 inches, 9 X-ray plates (12
x 16 inches), scanned at 300dpi: Size of scanned mosaic: 8-bit BW
125MB, 8-bit Colour 375MB, 16-bit BW 250MB and 16-bit Colour 753MB.
X-rays are generally mono or black and white so there is no real
need to store colour version of your scanned x-rays.

Basically, for paintings useful x-rays can be achieved by scanning
your X-rays to produce 300dpi 16-bit mono tiff images.

Processing and viewing these big images can also be complicated, but
there are several free pieces of software available that can handle
large 16-bit images, for example VIPS/Nip2
<URL:http://www.vips.ecs.soton.ac.uk/>. I hope this was useful,

Joe Padfield
Scientific Department
The National Gallery
Trafalgar Square
London WC2N 5DN
+44 20 7747 2553


                                  ***
                  Conservation DistList Instance 19:46
                 Distributed: Wednesday, March 22, 2006
                       Message Id: cdl-19-46-004
                                  ***