Image reconstruction (restoration) enhanced by invex regularizers has a wide application in vision tasks such as computed tomography, and magnetic resonance imaging. The use of an appropriate regularizer plays an important role in obtaining robust reconstruction results.
Imaging by invex regularizers is a process of reducing the complexity of a model through the inclusion of an additional parameter in order to reduce the overfitting of a model to the training data.
We develop algorithms for various types of imaging problems, such as:
- Tomographic 3D imaging
Our algorithms build upon linear algebraic, geometric, probabilistic and neural / deep learning operations.
Additionally, we attempt to develop theoretical foundations of the effect and principles underlying our algorithmic approaches.
Our current image restoration projects are mainly focused on the following directions:
- To seek effective algorithms addressing the deficiency of large-scale, noisy data with high redundancy, missing and small data situations, inadequate labeling cases.
- To improve downstream tasks like deep learning for imaging, and to provide more robust image reconstruction algorithms.
- To understand and mitigate the trade-off between model robustness and accuracy by both theoretical and empirical studies.
In addition to general image, we are particularly interested in issues and challenges identified in science and engineering domains, addressing issues like high-noise, missing information, data incompatibility, large scale, and memory usage.