Study On CT Image Reconstruction Algorithm Based On BM3D Algorithm And Gradual Projection Selection

Computed tomography technique has made a revolutionary impact on medical diagnosis, it can obtain the projection data from different angle when X-ray pass through object, then reconstruct the internal organization image using reconstruction algorithm. However, in practical applications, due to the limitation of radiation dose or the condition of the patient, it is not always possible to acquire complete projection data, and the reconstructed image has severe artifacts and noise in this condition. This dissertation is focused on how to reconstruct a high quality image when the projection data is insufficient.Recent years, compressed sensing theory has a rapid development, it provides a new potential direction to solve the insufficient projection data problem in CT field which can be regarded as using sparse sampling data to reconstruct original signal problem. Compressed sensing theory consider that if an original signal can be represented in another transform domain, then the signal can be exactly reconstructed from far less sampling data than Nyquist sampling theorem. By studying how the BM3D filter effect on the CT image, this paper proposes a reconstruction algorithm based on BM3D filter and algebraic reconstruction technique. Experiments show that the method can reconstruct the CT image from incomplete projection data effectively, and obtain a higher quality image compare with the total variation minimization method. Further, by analyzing the advantages of this BM3D method and total variation minimization method when the projection data is further reduced, this paper propose a iteration reconstruction algorithm based on BM3D filter and total variation minimization, experiments show that this method can obtain a better image.By studying the measurement matrix in CT reconstruction field, a reconstruction method by selecting projection data gradually is proposed. In this method, a low quality reconstruction image will be obtained from a small amount of projection data firstly, and then select the projection data based on the structural information, then update the reconstruction image to lead the next selecting process, and finally makes the image data achieve numerical convergence. This ray selection method provides an alternative option to select projection data in one projection angle, experiments show that this method can obtain higher quality reconstruction image than other selection method.At last, we make a summary and do a outlook for CT reconstruction method.