Research On Fast CT Iterative Image Reconstruct Based On Kohler Projection

In recent years,digital technology has achieved rapid development and application in the research and protection of cultural relics.Traditional three-dimensional scanning modeling cannot manifest the damage and structure of cultural relics.Computed Tomography(CT)technology can reconstruct the internal structure of cultural relics with nondestructive testing technology.The key point of CT technology is image reconstruction algorithm.Algebraic Reconstruction Techniques(ART)is a kind of iterative reconstruction algorithm,which is suitable for image reconstruction in the condition of incomplete projection.The calculation of projection coefficient has a great influence on the quality of reconstruction.Siddon algorithm is commonly used in ART algorithm to calculate projection coefficient.Because Siddon projection uses nearest neighbor interpolation,the reconstruction quality is not ideal.The K(?)hler projection method uses linear interpolation and Simpson integral to calculate coefficients,and access high reconstruction accuracy,but the K(?)hler algorithm has heavy calculation consumption and slow calculation speed.To solve this problem,fast image reconstruction based on K(?)hler projection is realized by improving projection algorithm and multi-core parallel acceleration.The main contributions of this thesis is as follows :(1)A fast fan-beam iterative CT image reconstruction method based on K(?)hler projection for 2D is proposed.By reducing the calculation of invalid crosspoint and the time-consuming of merging arrays,the crosspoint coordinates across the next grid are calculated rapidly by using the position relationship between rays and pixels.The crosspoint and midpoint pixel values are calculated by using linear interpolation,and then the projection coefficient is calculated by Simpson integral formula.The experimental results show that the fast K(?)hler method is about three times faster than the traditional K(?)hler algorithm while maintaining higher reconstruction accuracy.Meanwhile,the K(?)hler method achieves higher reconstruction quality than the Siddon,Joseph and DDM methods.(2)Based on the optimization of 2D point of intersection calculation,a fast cone-beam iterative CT image reconstruction method based on K(?)hler projection is proposed.In 3D condition,the crosspoint pixel value need to be calculated with the ray through bilinear interpolation,and the midpoint pixel value need to be calculated by using the trilinear interpolation.Finally,the projection coefficient is obtained and separated by Simpson integral formula.The experimental results show that the 3D fast K(?)hler method is about three times faster than the traditional K(?)hler method while maintaining higher reconstruction accuracy.(3)A parallel fast iterative CT image reconstruction method based on multi-core acceleration is proposed.By using the p Threads thread library,at the same angle and layer,the task is divided according to the number of rays on the detector plane,and the multi-thread parallel programming technology is realized to improve the reconstruction speed.In order to balance the calculation tasks of several thread,the number of voxels intersected with the ray at the same layer of the current angle is calculated in advance,and the calculation task is divided equally according to the total number of intersecting voxels.The experimental results show that compared with the serial ART algorithm,the four-core parallel algorithm achieves 3.1 times acceleration effect,and the six-core parallel algorithm achieves 4.4 times acceleration effect.