Cone-beam CT can use the flat panel detector to obtain the2D projection,and gettomographic images of objects through three-dimensional reconstruction. Compared with thetraditional-2D CT, It has some advantages, such as fast scanning speed, high utilization ofradiation, but the image quality is less than the two-dimensional CT. To solve this problem, Ipreprocess the two-dimensional projection datas, in order to suppress the noise whileimproving the quality of the reconstructed image. The main contents include:1. A variety of algorithms of CT image reconstruction are studyed, derive filteredback-projection algorithm: FBP and FD. I analyze the impact on image quali ty of theclassical algorithm FDK, which can be caused by the size of the cone-angle. Meanwhile, Iachieve some simulation experiments.2. It analyzed that the impact factors of CT image reconstruction, described the reasonsof artifacts detailly, such as the incomplete-data, beam hardening, Compton scattering and soon. It can provide the theory basis for projection data pre-processing.3. It was used that a projection data preprocessing algorithm combined by Bayes Shrinkthreshold and Window Shrink threshold based on wavelet transform. We all know that thewavelet transform has the characteristics of multi-resolution. Through the wavelet multi-scaleanalysis and coefficient control, I present a combination method of the wavelet and domainfiltering feature in order to make noise reduction for projection datas. Experimental resultsshow that the quality of reconstructed-images by noise reduction significantly improved andthe noise of reconstructed-images is effectively suppressed. |