CT Image Reconstruction Algorithm Research Of Local Region Tomography

Computed tomography (CT) is a technique of ascertaining objects internal structure onlythrough obtaining the external projection data but not destroying the object structure. It hasbeen widely used in diagnostic medicine and industrial non-destructive testing. The quality ofreconstructed image is related to whether the projection data are sufficient or not. However, itis not be able to obtain sufficient projection data in practical applications. The projection dataare truncated and limited because of considering the radiation dose, imaging system designand object structure. In fact, we only want to reconstruct the region of focal locations, not thewhole section image in medical diagnosis. In this paper, local CT reconstruction algorithmsfor region of interest are mainly studied. The specific contents and works are organized asfollows:Firstly, a local CT reconstruction algorithm is proposed based on wavelet analysis. Thenon-local property of traditional filter back-projection algorithm is analyzed. The propertiesof time-frequency local property, multi-resolution, high order vanishing moments and so on,are used to develop the wavelet CT reconstruction. Reconstruction error can be reduced byextrapolating the projection data and selecting suitable wavelet base function. Simulationshows that better results are obtained through wavelet reconstruction.Secondly, a total variation minimized algorithm is improved based on the compressedsensing theory. Under deep analysis, the total variation minimized algorithm is improved byusing conjugate gradient method instead of the ART reconstruction step, correcting the imageby non-negativity and minimizing the total variation, meanwhile introducing an acceleratingalgorithm. The reconstructed image is improved both in quality and speed.Finally, a local CT reconstruction algorithm is proposed based on wavelet frame.Because of the redundant and sparse represent property, a high-accuracy image can beobtained by penalizing the L1norm. This method is to solve the projection equation usingconjugate gradient method first and then shrinking wavelet frame coefficients in SplitBregman iteration. Simulation results show that the proposed algorithm works effectively.