研究目的
To propose a new joint model for the reconstruction of tomography data under limited angle sampling regimes, aiming to address artefacts by inpainting the missing data simultaneously with the reconstruction.
研究成果
The joint inpainting and reconstruction framework consistently recovers cleaner and more accurate structural information than current state-of-the-art methods in limited angle tomography, as demonstrated on synthetic and experimental datasets.
研究不足
The model's performance is sensitive to the choice of parameters and requires careful tuning. The numerical algorithm may converge to local minima due to the non-convex nature of the problem.
1：Experimental Design and Method Selection:
The study employs a joint model for tomography data reconstruction under limited angle sampling, utilizing a non-convex and non-smooth functional minimization approach.
2：Sample Selection and Data Sources:
Two synthetic datasets and one Electron Microscopy dataset are used. The synthetic datasets are discretized at a resolution of 200 × 200, simulated using the X-ray transform with a parallel beam geometry. The experimental dataset was acquired with an annular dark field (parallel beam) Scanning TEM modality.
3：List of Experimental Equipment and Materials:
MATLAB 2016b for implementation, PDHG algorithm for solving sub-problems, MOSEK solver via CVX for another set of sub-problems.
4：Experimental Procedures and Operational Workflow:
The algorithm involves alternating descent steps for minimizing the joint functional, with parameters chosen based on the dataset.
5：Data Analysis Methods:
The performance is evaluated based on the ability to recover cleaner and more accurate structural information compared to current state-of-the-art methods.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
