Research On Industrial Cone Beam CT Iterative Reconstruction Algorithms And Artifact Correction

In Industrial Cone Beam Volume Computed Tomography (CBVCT) system, there aremany factors that could affect the iterative reconstruction result. One of the main factors isweight factor, including its weighted model, computing method, calculated quantity, and theexecution efficiency of corresponding program. So, how to calculate weight factor quicklyand efficiently is very important to iterative reconstruction algorithms. After determiningthese factors, different iterative algorithms still have different reconstruction results even withthe same reconstruction parameters. Therefore, choosing an appropriate algorithm accordingto the reconstructed request also is very important. What’s more, the CBVCT system iscomplex, any problems of hardware and software during collecting projection data wouldresult in some image artifacts and lower image quality, so, choosing an appropriate correctionmethod to remove artifacts and improve reconstruction image quality is necessary to theassessment of the quality of detected object.Based on these questions, this paper mainly researches the iterative algorithms inCBVCT system and the correction of artifacts in reconstruction image. The main researchcontent is as follows:1. Research the principle and the procedure of ART detailedly, analyze factors ofiterative algorithms that could affect the reconstruction result. Aim to the key factor, weightfactor, a new method is improved to calculate it quickly and accurately. By comparing theexecution efficiency of proposed method and some referenced methods reconstructingexperimental data, demonstrate that the proposed method is more quick and the reconstructionimage also has a good quality.2. Based on the research of ART, other iterative algorithms, SART, MLEM are alsoresearched in detail. What’s more, the ordered subset is applied to SART and MLEM, to accelerate convergence rate of reconstruction image. The real projection data arereconstructed separately byART, SART, MLEM, OSSART and OSEM, and the reconstructionimages are compared to discuss properties of every algorithms.3. Aim to the streaking and ring artifacts occurred in the reconstructed solid rocketengine simulator, the reasons are analyzed detailedly from projections. Because of somedisadvantages of normal used correction methods, a new method is proposed to correctartifacts. The new method is using morphological opening and closing operation to processprojection data, calibrating bad pixels of FPD, reducing noise of projection data. As a result,the artifacts in reconstruction images are eliminated. The experimental results show that theproposed method is simple and efficient.