Research On Data Correction In Computed Tomography

In Computed Tomography (CT) system, projection data often become abnormal, due to detected objects, scanning background and other factors. These nonideal data for example, metal projection with huge value or projection with high level noise, bring in many artifacts or noisy points in CT images and greatly damage the image quality. In order to improve the quality of image, many algorithms have been proposed and applied for correcting projection data and therefore the reconstructed image can become able to accurately depict the structure of the cross-section of the object, which has significant meaning for CT system's performance on medical imaging and industrial non-destructive test.With the intent of improving the quality of CT images, this thesis begins with the introduction of many established correction methods and discusses the capabilities of several interpolation manners in the interpolation-type correction method. The thesis pays more attention on the research of metal artifacts reduction (MAR) in conventional filtration backprojection (FBP) reconstruction. Furthermore, the PI-line segments based backprojection filtration (BPF) algorithm has been studied; several factors relative to reconstruction effect and metal artifacts have been well discussed. Also, de-noising CT images have been practiced via the Partial differential Equation (PDE) method.In the first part, the existed methods for MAR have been analyzed. Since interpolation has been broadly applied for MAR in CT system and many modified correction methods derive from interpolation, the performance of several general manners of interpolation have therefore been compared and discussed.The second part is the focus of the thesis. According the specific requirement of MAR, several effective correction methods have been proposed, including accurate identification of metal projection based interpolation method, wavelet multi-resolution interpolation method, projection nonlinear attenuation method and Euler's elastica and curvature based sinogram inpainting method. Each methodbegins with the background, and then the algorithm's procedure and the corrected results are provided. The conclusion or discussion goes at the end.Concerning the BPF algorithm which has developed for only several years, the simplified parallel PI-line segments based parallel beam BPF algorithm is introduced, and factors relative to reconstruction effect are also discussed. In the end, the metal artifacts are studied and corrected.The thesis has also made an effort on CT image denoising. Different with other image, CT image's noise derives from many factors. In this part, we study the sources of CT image's noise and the PDE denoising method. In addition, we apply anisotropic PDE method to denoise CT images and the denoised results are discussed.