During the summer of 1994 at Los Alamos National Laboratory (LANL), I produced a code, SYN_IMG3, to generate synthetic radiographs. Using SYN_IMG3, I analyzed the accuracy of tomographic reconstructions by the LANL Hydrodynamics group. The current reconstruction algorithm makes the simplifying assumption that photons exposing the original image are monoenergetic, when in fact they are not. I showed that this introduces errors of 5-10%, and advocated a new treatment incorporating the distribution of incident photon energies. My work was conducted under the supervision of Scott Watson and Karl Mueller, both research scientists in Hydrodynamics.
Radiography helps understand how an object behaves while under stress. For example, an image of an explosively driven metal object can show how the metal flows like a fluid. Simulating the radiographic process allows computation of synthetic radiographs. This technique allows study of how different exposure parameters affect image quality, and is useful both for designing new radiographic facilities and for finding an optimum configuration for a "real-life" exposure.
Tomographic reconstruction permits deducing the 3D state of an object from one or more 2D radiographs. Synthetic radiography generates an ideal image with no noise or blur, and can isolate errors introduced by the reconstruction. The current algorithm assumes X-ray photons exposing the image are monoenergetic, when in fact they exhibit a Bremsstrahlung spectrum of energies. Error introduced by this simplification was believed minor compared to the 5-10% errors from film blur and noise. But my treatment isolated the error due to the monoenergetic assumption and showed its magnitude is in fact significant, especially in light of a new facility under construction with improved noise and blur characteristics.
This work therefore demonstrates that spectral effects do play an important role in tomographic reconstruction and need to be incorporated. It also shows that synthetic radiography should see wider use, both for designing experiments and for understanding the limitations of current facilities and techniques.