Image Quality in Digital Industrial Radiography,
U. Ewert, U. Zscherpel, K. Heyne, M. Jechow and K. Bavendiek

Material Evaluation, vol. 70, no. 8, pp. 955-964, 2012

Hi Klaus:

Could you please recommend some textbooks about DR, especially have wide and deep discussion of PSF and MTF?

Colin, in industrial DR the MTF is not used very often; when we did the ASTM meetings to decide how we want to measure the spatial resolution in DR we decided to take the Duplex Wire. The result with this method is directly visible and the users would not need complex mathematical functions to get a value.
For evaluation of CT systems the MTF with the surface of a cylinder is used (E1695-20). There is a guide in ASTM (E1441-19 Guide for Computed Tomography (CT)) which describes on pages 9 and 10 the mathematical background and usage of this function in a CT system.

I had never seen a PSF in DR ... (to my knowledge, only edges are evaluated with the ESF -> LSF -> MTF). The PSF is a 2D function and after the Fourier transform you would have a two dimensional optical transfer function; in DR (and CT) we would only use the spatial resolution in one direction (the one dimensional optical transfer function is the MTF).

(I think you already had a look into Wiki for and for ).

Hi Klaus:

How can we measure defects' size as precisely as possible in DR?

I probably got different results when I measured both the pixel number and pixel size repeatedly.

Hi Colin,
best would be to place an object of know size beside the defect you want to measure.
I used to take an ASTM E1025 or E1742 plaque hole IQI, and put it on the object in the beam at the position where I located the defect.
Then I used the line profile function in the YXLON software Image 3500 and calibrated the pixel size by the IQI dimension at the position of the defect:

The E1742 penny is 2" long (50.8mm or 50800µm for higher precision)

In this example I simulate the defect with the 4T hole of the IQI. For the hole size I read in ASTM E1742: Minimum diameter for 4T hole = 0.040 in. which is 1.016mm (or 1016µm).
I open the line profile function in the YXLON software, make the Subpixel function active and set as start point the left side of the hole (green arrow):

Now I move the mouse pointer to the right side of the hole and see the dimension 1039.1µm (which is quite near to the nominative 1016µm)

I know that measurement in subpixel range is very sensitive on noise ... (but offers the highest resolution for measurement in the image)