Cone-Beam Computed Tomography Deblurring Using an Overrelaxed Chambolle-Pock Algorithm

The scatter contamination and photon starvation artifacts are less severe due to the smaller cone angle and attenuation associated with the smaller sample size in the microstructure imaging (MSI) with a cone-beam CT (CBCT) instrument. Compared with other artifacts, the blurring effect is more critical in the MSI due to the direct degradation of spatial resolution. An efficient deblurring method using the line integral images is proposed in this paper. The point spread function (PSF) of the CBCT system is simplified as a 2D Gaussian kernel function with the spatially invariant assumption. The deblurred line integral images are calculated using the iterative optimization process with an objective function in which the blurring effects are described by the convolution between the simplified PSF and the real line integral image. A first-order primal–dual algorithm is applied and derived to solve the real line integral image due to its fast convergence and high computational efficiency. The performance of the proposed method is evaluated using various datasets, including a digital phantom, a physical phantom, and a laboratory mouse. The spatial resolution, noise distribution, and computation cost of the proposed method are compared with those of a 3D image domain deconvolution method. In addition to the well-suppressed blurring effects in the CBCT images, the proposed method provides a higher computational efficiency than the 3D approach. The proposed method is thus practical and attractive to be incorporated into the data processing workflow of the CBCT instrument of MSI.