Thesis/Dissertation Guide For Postgraduates Of XIDIAN UNIVERSITY

The multi-layers ball(MLB)is a kind of high-precision device and its physical performance depends strictly on the material quality,machining accuracy and assembly accuracy.Before being used,the MLB should be checked carefully and confirmed to be normal.Due to its unusual structure,big volume and density diversity among the spherical shells,there is still no available method to check a MLB without detaching it.Although high-energy X-ray tomography is one of the possible methods that can be used to test the MLB in the future,some special problems,caused by the special structure of the MLB,should be handled firstly,which include the problems in X-ray projection simulation,CT reconstruction from insufficient projection,discovering defects in different parts of a MLB through one image,and so on.This dissertation has been mainly focused on dealing with the key parts of the above problems,and consequently formed a method for testing the MLB.The author's major contributions are outlined as follows:1.A new method of X-ray projection simulation is proposed.Firstly,we model the testing objects by CAD software and convert the piece of interest in the model into a pseudo color or gray image,and then fill the different regions in the image with different material parameters by image segmentation and recognition method.To calculate the projection data,we use a rotation method to change the orientation of the image array and an integration calculation method to get X-ray path lengths in different materials.By the method,projection simulation for any testing object can be easily carried out with no need for complicated algorithm and program design.This work has prepared and provided an available tool for beginning the research on CT reconstruction method.2.A new reconstruction method named total variation constrained iterative filtered back-projection is put forward.Being limited by the solving theory of analytical algorithm,it is intractable to utilize priori information and constraint condition in CT reconstruction.To resolve the problem,an iterative method based on filtered back-projection(FBP)is proposed.To get a continuous improvement of the CT image quality,a total variation model is introduced into the iterative procedure at the same time.The results of numerical simulation indicated that the method is very suitable for high quality CT reconstruction,whether from ideal,sparse,metal or limited angle sinogram.The algorithm has been used as the core algorithm for the CT reconstruction of the MLB.3.The CT reconstruction method for MLB is studied.According to the structure of MLB and the characters of its sinogram,an object-oriented CT reconstruction method is put forward.It is suggested that the insufficient projection of MLB should be repaired at first,and then the defects information and the background information should be separated from the sinogram secondly.At last,CT images for the defects and the background are reconstructed respectively from the separated sonograms.Tests showed that this method can effectively limit the affection of the insufficient projection on the CT image of MLB.At the same time,because the information of minute defects can avoid being disturbed and flooded by the very background information in the reconstruction process,the image quality has been improved significantly.4.A dual-energy CT is designed virtually for testing MLB which consists of a high Z core embraced by low Z materials.According to this method,the high-density parts and the low-density parts in the MLB are projected by X-ray beams of different energy,and then the CT images are reconstructed from different sonograms and fused into high-quality CT image for MLB.The Method for X-ray energy selection,the design for the X-ray projection system and the algorithm for dual-energy CT reconstruction are illustrated in this thesis.5.To remove the impulse noises in radiography,a joint denoising method based on the grayscale morphological reconstruction and the improved non-local means(NL-means)filter is presented.In this method,generalized filters are abandoned,so that details in images can be kept perfectly.Experiments showed that impulse noises in various radiographs can be removed effectively by the method.Furthermore,when the joint method was combined with the classical NL-means method and then used to remove the noises in X-ray sinograms,an obvious improvement of the CT image was received.6.A preprocessing technique is proposed for the correction of ring artifact in X-ray CT images.By fitting projections in the sinogram with piecewise polynomial,a set of candidate correction factors for mis-calibrated detector are created and then the factor with maximum probability occurrence is picked out to rectify the projection data.In this thesis,principle and procedure of the method is illustrated.Tests under different condition showed that sparse or continuous ring artifacts in CT images can both be rectified effectively,while spatial resolution of the images preserved.CT images for a multi-material object can also be corrected by the method.