Abstract:We study numerical methods of tomography in domains with a reflecting obstacle. It will be shown that tomography with sets containing both broken rays, i.e. rays reflecting at the obstacle, as well as unbroken rays, has a smaller error between the original and reconstructed image compared to classical tomography methods.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to improve the image reconstruction accuracy and numerical stability of tomography by numerical methods in the presence of reflective obstacles. Specifically:
1. **Problem Background**:
- In fields such as biology, geophysics, oceanography, and communications, it is necessary to infer information about structures such as the human body, the interior of the earth, the ocean, or the atmosphere from measurement data at the domain boundary.
- When there are reflective obstacles in the domain, traditional tomography methods will encounter new challenges because the signal may be reflected by the obstacles during propagation, thus affecting the reconstruction results.
2. **Research Objectives**:
- Research tomography methods that include broken rays and unbroken rays to reduce the error between the reconstructed image and the original image.
- Prove that in the presence of reflective obstacles, using broken rays can improve the accuracy and numerical stability of image reconstruction.
3. **Main Contributions**:
- Propose a numerical method for tomography using broken rays in the presence of reflective obstacles.
- Verify through numerical experiments that the tomography method combining broken rays and unbroken rays has a smaller reconstruction error than the traditional method using only unbroken rays.
4. **Theoretical Basis**:
- Based on the theory proposed by Eskin et al., the paper proves that in the presence of reflective obstacles, the tomography problem is well - posed, that is, there is a unique solution and the solution is stable with respect to changes in the input data.
- This theory provides a solid mathematical foundation for developing numerical methods.
5. **Numerical Methods**:
- Use the Kaczmarz method (a method for iteratively solving linear equations) to solve the tomography problem including broken rays and unbroken rays.
- Through numerical experiments, it is shown that in the presence of reflective obstacles, the tomography method combining broken rays can significantly improve the quality of the reconstructed image.
In summary, the main purpose of this paper is to solve the image reconstruction problem in tomography in the presence of reflective obstacles by introducing broken rays, and to verify its effectiveness and superiority by numerical methods.