Browsing by Author "Bradley, Hamish"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Sub-pixel registration for low cost, low dosage, x-ray phase contrast imaging : a thesis presented in partial fulfilment of the requirements for the degree of Master of Engineering in Electronic and Computer Engineering at Massey University, Palmerston North, New Zealand
    (Massey University, 2021) Bradley, Hamish
    X-ray phase contrast imaging is an imaging modality that measures the phase shift of the X-ray wavefronts as they travel through different materials. This gives a higher contrast between regions of similar X-ray attenuation, in a medical sense this corresponds to a higher contrast of soft tissues. A new area of research for X-ray phase contrast imaging is to use shadow-based intensity modulation to generate these images. This thesis explores a range of different registration techniques, and their suitability for phase contrast imaging using shadow based intensity modulation. Image registration is a key step in generating the phase contrast images as it is related to the x and y differential phase. These are then integrated to generate the phase contrast image. Therefor a high accuracy sub-pixel registration technique will provide high quality phase contrast images. The registration techniques explored are 1D curve 2-D surface fitting to a correlation map, phase registration, Newton-Raphson method, and optimal interpolation filtering. These registration techniques were tested with images that are noise free, as well as images corrupted by Poisson noise. The Newton-Raphson, and the optimal interpolation filters show the most promise due to low errors in the noise free environment. In the presence of noise, the Newton-Raphson method performs poorly, and hence requires a good denoising method, while the optimal interpolation filters do not get any improvement from any denoising techniques. Currently the Newton-Raphson based method are used widely in digital image correlation, however the optimal interpolation filtering has the benefit of not being limited by the choice of interpolation technique, and it removes the iterative process, and depending on the size of the optimal interpolation filter it performs better than, or only marginally worse than the Newton-Raphson method.