Research On Removing Ring Artifacts In CBCT Images

In recent years, cone-beam computed tomography has been widely used in medical field such as clinical diagnosis, 3d implants and so on. As one kind of computed tomography, cone-beam CT has precise imaging procedure and efficient utilization ratio of rays. However, because of the limit of technological design and reconstruction methods, cone-beam CT always has a series of concentric rings which have the same center and different intensity called the ring artifact. The appearance of ring artifact seriously influences the quality and facticity of images and interferes the clinical diagnosis. Therefore, removing the ring artifacts and keeping the image resolution at the same time has become the research focus of academic circles.In this paper, we propose a new post-processing technique which could remove the ring artifacts effectively. Meanwhile, it could keep the details and edges of the image clearly.At first, in order to change the geometric characteristic of ring artifact and reduce the difficulty of processing. We transform the image from cartesian coordinate system into polar coordinate system where the ring artifacts perform as the stripe artifacts. And then, we adopt the 0 smoothing to process the CT image. The proof of 0-norm is decomposed into two subproblems which have closed-form solutions. According to the superiority of 0 smoothing, the main details and edges of the CT image will be reserved. About the compensation of details, this paper adopts the artifact template which is extracted by the median value. OSTU multi-threshold segmentation which is used to extract region of interest could eliminate the limitation of 0 smoothing. Median value extraction in this region of difference image between origin image and filtered image could guarantee the accuracy of artifact template. Finally, we subtract the artifact template in the difference image, compensate the details into filtered image and transform the image back to cartesian coordinate system.Finally, experiments which use the algorithm in this paper are done based on multiple cone-beam CT images and compared with other kind of algorithm. Experimental results show that the algorithm based on 0 smoothing and OSTU multi-threshold segmentation is superior to other algorithm for the removal of ring artifacts in cone-beam CT images.