|
|
Noise reduction in high resolution computed tomography data
Tkadlecová, Markéta ; Chmelík,, Jiří (oponent) ; Zikmund, Tomáš (vedoucí práce)
Due to the complexity of the computed tomography (CT) data acquisition process, the noise in captured X-ray images is inevitable and can distort acquired information. Therefore, the noise should be controlled. Noise reduction in CT data has primarily been studied in low-dose medical CT, and there is little known about the properties of noise in sub-micron CT and its suppression. The denoising strategies can take place in the X-ray images, CT slices, or during the reconstruction process. This work primarily focuses on the reduction of noise in X-ray images and CT slices. The first step to finding a complex denoising methodology is to determine the distribution and mathematical model of the noise in the X-ray images. The model was established using bright field images taken under different exposure times. With the estimated model, the artificial noise in a phantom dataset could be simulated. Selected algorithms were tested in the X-ray images and the tomogram slices of the phantom and compared subjectively and objectively, through visual inspection and image quality metrics. The denoising strategies with the best outcomes were further evaluated on measured submicron CT datasets from the CT system Rigaku nano3DX, and their advantages, limitations, and possible usage were described.
|
|
|
Material Decomposition in Spectral Submicron Computed Tomography
Mikuláček, Pavel ; Zikmund, Tomáš (oponent) ; Štarha, Pavel (vedoucí práce)
Computed tomography is a non-destructive method for imaging internal structures of samples. It is mainly used in medicine, but also in industry or scientific fields. Spectral computed tomography is also commonly used in medicine, which allows the differentiation of materials based on their attenuation properties, which are dependent on the energy of the X-rays. This approach is also beginning to appear in industrial applications. This thesis deals with dual-energy computed tomography, which is a specific form of spectral tomography using two different X-ray spectra. Material differentiation, or decomposition, can be performed either on projection data or on their reconstructions. The output are images containing information about the concentrations of individual materials found in the scanned volume. In addition to the concentrations, the atomic numbers and densities of these materials can also be determined. In this thesis, a basic method of material decomposition was implemented, the outputs of which serve as initial guesses in a series of iterative algorithms that perform the decomposition. A method has also been implemented that reduces noise in decomposed images whose values are highly correlated. The algorithms were tested on artificial data, but also on real data scanned by the Rigaku Nano3DX device. A comparison of the effect of various tomographic artifacts on the resulting material decomposition was made.
|
host ::
RSS.