Joint Regularized-based Image Reconstruction by Combining Super-Resolution Sinogram for Computed Tomography Imaging

In computed tomography imaging, the $2 \times 2$ acquisition mode improves the projection collection efficiency and reduces the X-ray exposure time; however, the collected projection is low-resolution and the reconstructed image quality is poor. Although the super-resolution (SR) method can improve the quality of the acquired projection in $2 \times 2$ acquisition mode, the signal-to-noise ratio of the reconstructed image is still affected by the estimation errors between the SR sinograms and the high-resolution sinograms. In this study, a joint regularized-based reconstruction method was proposed. Under the condition of obtaining SR sinograms, we utilized the system matrix in $1 \times 1$ and $2 \times 2$ projection acquisition modes to construct the fidelity terms. In addition, the block matching and total variation regularizations were used to fully depict the sparsity of images. The proposed reconstruction model was solved by the iterative alternating minimization method. The experimental results on real anthropomorphic phantom data show that the proposed method is capable of suppressing noises while maintaining image details, which are not observed in the reconstructed results of other compared methods.