Research On Key Techniques Of CT Image Reconstruction

CT (Computerized Tomography) is an advanced technique mixed several subjects developing quickly these years. Besides the medical diagoses,it is still used many important fields such as industrial non-destructive detecting, radio astronomy ,etc.CT image reconstruction play the core role in CT technique ,the performance of reconstruction algorithm is of vital importance to the judgment of detected results. Analytical algorithm was deduced based on rigorous mathematic theroys, and have the advantage such as smaller memory consumption and shorter running time compare to the ATR kinds. So the analytical algorithm is used widely on CT detection equipment. The market demand have influence on the development of technologe.with the aim of image resolution improvement, It is the unremitting developing goal to improve the classical algorithm, introduce new method.today the higher application such as inner precision non-destructive in industrial and converse engineering bring forward the higher demand to CT reconstructed performance. Three dimension imaging become the developing direction in future. But it is facing the difficulties consist of plenty of data and complex computing, which are restrict the usage of 3D CT in wider areas. So speedup the 3D reconstruction is a research hotspot. In the case of projections can not be acquired adequately under some special situations, how to reconstruct the image using the inadequate data become an attractive and hard matter.Aimming at the problem mentioned above, the paper achieve the following fruits.FBP (Filtered Backprojection)is the main algorithm in two dimension image reconstruction. It is filter function that has a close relationship with the quality of image reconstruction. The Gibbs phenomena will emerges when the filter function truncated by ideal rectangle window. The paper introduces a method to eliminate Gibbs by Fejer core function based on C-1 principle in infinite series sum theory. The experiment result shows the method is more efficient comparing with the ways of adding RL window and hamming window. The paper deduces a DHB algorithm based on FBP and Radon reverse transform. The algorithm uses the symmetrical property of X-ray in fan-beam under 360 scanning to eliminate the distant weighting function which has the relationship with the change of scanning angle and pixel location by bring in a ratio factor reflecting the value of pixel contributing .The algorithm avoid the instability when the weighting function has a small value. At the same time, the formula becomes simple which make it possible to speedup the computing speedup the reconstruction much more, the paper proposes a software method .The way is arrange the fan-beam projections to parallel projections, and then backproject in polar coordinate. The minimum scanning scope and the minimum available data are given after analysis of projection redundance, so a short scanning mode can be done to save the scanning time. When backproject in polar coordinate, we use a method called MSBP to decrease the compute time of pixel location .here are plenty of projections and complex calculations in 3D reconstruction. It is hardware solution that can be efficient to satisfy the real-time demand. The paper considers that FPGA is the most appropriate hardware on which the FDK algorithm run following the compare with others hardware such as GPU, workstation.Cell and so on. We design a circuit to implement FDK backprojection in eight path synchronously using Xilinx xc5vlx110.The simulation result shows the speedup ratio arrive at 252.Aiming at the problem of limited-view reconstruction, the paper makes a primary research. A new method belong to analytical kind is proposed. Following the original reconstruction, the FOI (field of interesting) is located, and then fill the missing projection on sinogram according to the acquired projections, finilly reconstruct the last image utilizing the whole projections. The method is simple to realize compare with iterative kind, and the image reconstructed is good when inner structure of detected object is regular.