The Study On The Image Reconstruction Algorithm Of Industrial CT By Computer Simulation

In this paper we studied filter backprojection algorithm and algebra iteractive algorithm of industrial CT image reconstruction. We analyzed the main factors that influence on the quality of reconstruction image. These factors are filter function, projection data, sampling space and noise level in filter backprojection, and the choice of relaxation parameter, the fashion of data access, the choice of basis function and the estimate of optimum iterative times in algebra iterative algorithm. Then we modified the RL filter function by multi-point average method, that is, MRL filter function. It is found that MRL filter function is better than RL filter function in restraining noise and eliminating false trace. In order to look for a balance between material cost and the quality of reconstruction image, we simulated how sample space and projection data influence on the quality of reconstruction image. Then we draw a conclusion: the quality of reconstruction is more sensitive to sample space than to the number of projection in noise level; the quality of reconstruction would drop if the sample space is too narrow in high noise level. For the sake of the image reconstruction efficiency we shortened the reconstruction time almost 100 times by parameter table method. In algebra iterative algorithm we found the range of relaxation parameter is very different. The range of relaxation parameter of ART was narrower than that of SIRT. Different methods of data access will influence on the quality of reconstruction image. We presented a criterion for the estimate of optimum iterative times. It can help us to aviod unuseful iterative.In this paper we simulated how the different part of energy spectrum influence on the quality of reconstruction image by MCNP4B program without shield and with Pb material shield. It is found that there is little influence on the quality of reconstruction image when the different part of energy spectrum is used as projection data. We use surface source card to sample instead of sphere source, as a result the sample time is much shortened .At last we analyzed the image false trace during the course of reconstruction. The false trace can be depressed by Matlab image technique.We program software and make simulative experiments. We compare the filter backprojection algorithm with algebra iterative algorithm and draw some useful experience. It would help us make better in CT.