<p>Low-dose computed tomography (CT) is crucial in medical imaging as it minimizes radiation exposure to patients. However, reducing the radiation dose often leads to increased noise and artifacts in reconstructed images, hindering accurate diagnosis. To address this, we propose a novel low-rank tensor image reconstruction model that effectively exploits the multi-scale and multi-directional information inherent in a single CT image. Our model leverages the nonsubsampled shearlet transform (NSST) to construct a 3D tensor composed of the shearlet coefficients and the repeated instances of the CT image. This innovative coupling of shearlet and spatial domain information allows for effective noise suppression while preserving crucial structural details. A low-rank constraint is then enforced on the constructed tensor using the tensor singular value decomposition algorithm. The final reconstructed image is selected from the tensor based on its denoising performance. Extensive simulations and real data experiments demonstrate that our method surpasses comparative algorithms in objective evaluation metrics and subjective visual quality. This approach enhances the quality of low-dose CT images, improving the reliability of medical diagnoses.</p>