Institution
Georgia Southern University
Abstract
Computed tomography (CT) has been widely applied in medical imaging and industry for over decades. CT reconstruction from limited projection data is of particular importance. The total variation or l1-norm regularization has been widely used for image reconstruction in computed tomography (CT). Images in computed tomography (CT) are mostly piece-wise constant so the gradient images are considered as sparse images. The l0-norm of the gradients of an image provides a measurement of the sparsity of gradients of the image. However, the l0-norm regularization problem is NP hard. In this talk, we present two new models for CT image reconstruction from limited-angle projections. In one model we propose the smoothed l0-norm and l1-norm regularization using the nonmonotone alternating direction algorithm. In the other model we propose a combined l1-norm and l0-norm regularization model for better edge preserving.
Sessions
Two Regularization Models for Computed Tomography Image Reconstruction from Limited Projection Data
Computed tomography (CT) has been widely applied in medical imaging and industry for over decades. CT reconstruction from limited projection data is of particular importance. The total variation or l1-norm regularization has been widely used for image reconstruction in computed tomography (CT). Images in computed tomography (CT) are mostly piece-wise constant so the gradient images are considered as sparse images. The l0-norm of the gradients of an image provides a measurement of the sparsity of gradients of the image. However, the l0-norm regularization problem is NP hard. In this talk, we present two new models for CT image reconstruction from limited-angle projections. In one model we propose the smoothed l0-norm and l1-norm regularization using the nonmonotone alternating direction algorithm. In the other model we propose a combined l1-norm and l0-norm regularization model for better edge preserving.
