| The Industrial Computed Tomography(ICT) inspection need deal with large processing data and requests high resolution of reconstruction image, thus demanding high-speed of reconstruction algorithm. It always is a major project in the study of ICT to improve the timing performance of reconstruction algorithm without degradation in image quality. The main work of this paper is as follows:(1)We propose a fast filtered backprojection(FBP) algorithm, which is based on regular hexagonal pixels, for CT fan-beam involving collinear equispaced detectors. There is 12-symmetrical relation of projection address and weighted coefficient of pixels in regular hexagonal pixel subdivision, which means we can get the projection address and weighted coefficient of all pixels with only one-twelfth calculating of projection address and weighted coefficient of total pixels. Consequently, we can reduce redundant calculation and accelerate reconstruction speed. In this paper, we achieve parallel computing by using of SSE(Streaming SIMD Extension). The reconstruction speed is promoted nearly 1/3 times than the traditional Square-Pixel-based filtered backprojection algorithm. The validity of the proposed algorithm is verified by the real CT data.Further, the 12-symmetrical relation of regular hexagonal pixel subdivision is fit for GPU accelerating and the method of pixel subdivision can be adopted in multi-slice helical CT reconstruction.(2)We propose a fast regular hexagonal pixel-based parallel interpolative algebraic reconstruction (IART) algorithm. The main computation of IART is in computing projection address of pixels. We use the 12-symmetrical relation of regular hexagonal pixel subdivision in the algorithm to optimize calculating order and reduce redundant calculation. The results of experiment show that the reconstruction speed is largely promoted compared with the traditional square-pixel-based IART algorithm. |
PDF Full Text Request