Research On Local Reconstruction Algorithm Of CT Images

CT (Computed Tomography) is a kind of advanced nondestructive testing technology. In its many applications, it is common to meet the cases, such as the image reconstruction of region of interest, the image reconstruction of big object, the image reconstruction of high spatical resolution. These all can produce local reconstruction problems. The local reconstruction with truncated projection data is a difficulty in the field of CT reconstruction. Conventional reconstruction algorithms used to deal with the reconstruction of truncated projection data will product severe truncation artifacts. The image reconstruction of practical CT system is a course of complex computing of massive data. The exsiting local reconstruction algorithms have many flaws in reconstruction efficiency and parallel performance, which can’t satisfy the requirement of fast and accurate reconstruction for practical CT system. In this paper, aiming at the different concrete applications, we study and design the local reconstruction algorithms with good reconstruction efficiency and parallel performance for practical CT system in the case of the guarantee of reconstruction quality. The main work and innovation of this thesis can be summarized as follows.1. For some local reconstruction problem (no truncation along PI-line direction), the BPF-type local reconstruction algorithm based on data rebinning method is developed. It is performed by firstly rearranging the cone-beam data to tent-like parallel-beam data, and then applying a proposed BPF-type algorithm to reconstruct images from the rearranged data. In case of the guarantee of reconstruction quality, the reconstruction efficiency of the proposed method has an improvement over the original BPF algorithm. Moreover, there are no relativities in the implementation of the proposed method, so it is fit for parallel computing algorithm. The experiments of numerical simulation and real data reconstruction have demonstrated the correctness and advantages of the proposed algorithm.2. For the general local reconstruction problem, an approximate truncation resistant algorithm based on Radon inversion transform is developed. The algorithm extends the2D Radon inversion transform to the3D local reconstruction in the circular geometry. It achieves data filtering in two steps. The first step is the derivative of projections which acts locally on the data and can thus be carried out accurately even in presence of data truncation. The second step is the nonlocal Hilbert filtering. Not only the reconstruction efficiency of the proposed method is considerable like FDK algorithm, but also it has comparable ability to restrain truncation artifacts like the approximate truncation resistant algorithm for computed tomography (ATRACT). The numerical simulations and real data reconstructions have been conducted to validate the new reconstruction algorithm. The presented local reconstruction algorithm based on Radon inversion transform provides a simple and efficient approach for the approximate reconstruction from truncated projections in circular cone-beam CT.3. Aiming at the local reconstruction of special structural specimens which is plate-like, an improved BPF algorithm is presented to reconstruct the plate-like specimens in a super-short scan, thus reducing imaging time and increasing practical CT system throughput. The algorithm of the proposed method is an improved backprojection-filtration (BPF) algorithm using an integral operator with fixed integral interval. It is found that if the thickness of the reconstructed plate-like specimen is less than0.0349R (R is the scanning radius), the uncertainty of the proposed method can be ignored. When the thickness of the reconstructed plate-like specimen is a little large, it can be reconstructed by increasing the scanning radius befittingly. The results of numerical simulation and real data reconstructions show that the proposed method is a good choice for the reconstructions of plate-like specimens, because it can not only yield images with quality comparable to that obtained with existing algorithms, but also reduce imaging time and; increase CT system throughput.4. Aiming at the local reconstruction of special structural flat object, a method based on FDK algorithm is developed. It is found that if the thickness and the horizontal length of local area satisfy some conditions, accurate reconstruction can be carried out using efficient FDK algorithm. From the results of numerical experiments, an experimental condition is obtained. That is to say when the horizontal length of local region is bigger than the1/8horizontal length of the reconstructed object and the thickness of local region is less than the1/13horizontal length of the reconstructed object. The reconstruction results of numerical simulation and real data confirm FDK algorithm can reconstruct the local area of flat object well under the condition.5. For the image reconstruction problem that the reconstructed object is big and can not be covered by the field of view, we consider a scanning configuration in which X-ray beams only cover half of the object and the cone-beam projection data are acquired from an asymmetrically positioned half-sized detector. The acquired cone-beam projection data are truncated at every view angle, which does not satisfy the conventional reconstruction condition that the projection data can not be transversely truncated. If an explicit data rebinning process is not invoked, this data acquisition configuration will play havoc with many known cone-beam image reconstruction algorithms. To reconstruct images from the truncated projection data at any view, we apply a recently developed backprojection filtration (BPF) algorithm in circular cone-beam CT and an observation that a correct backprojection image can be formed by combining the projection data from different view angles, and then develop an algorithm to reconstruct3D images for the half-covered scanning configuration. Not only the proposed algorithm does not need the data rebinning process, but also the implementation of backprojection image just needs the projection data in the range of180degree for every reconstruction point. These mean that the proposed algorithm has good reconstruction efficiency and high spatial resolution. Numerical simulations and real data reconstruction experiments are conducted to validate the proposed reconstruction algorithm.Finally, we summarize our research work for the thesis, and discuss some research topics and directions relative to this work in the future.