Research
Research directions of the LIMO Lab.
LIMO studies the mathematical and computational foundations of imaging and data science. Our work connects rigorous analysis with algorithms designed for real measurement, reconstruction, and learning problems.
Inverse Problems and Computational Imaging
We develop mathematical models and reliable algorithms for recovering information from incomplete, noisy, or indirect measurements.
Our work combines variational modeling, inverse-problem theory, and modern computational methods to make image reconstruction and restoration more accurate, interpretable, and efficient.
Methods
- Variational and regularization methods
- Model-based and data-driven reconstruction
- Sparse and low-complexity modeling
Applications
- Image reconstruction and restoration
- Computational sensing
- Scientific and engineering imaging
Medical Imaging
We build reconstruction, correction, and analysis methods that connect mathematical insight with clinically relevant imaging problems.
We study imaging pipelines in which physical acquisition models, limited observations, and learned priors must work together. Current applications represented in the lab's work include CT, PET, SPECT, MRI, and X-ray image analysis.
Methods
- Physics-informed reconstruction
- Attenuation and scatter correction
- Representation and anomaly learning
Applications
- CT, PET, SPECT, and MRI
- Low-dose and limited-angle imaging
- X-ray image analysis
Optimization and Machine Learning
We study the algorithms and mathematical structures that make large-scale learning and data-driven computation efficient and dependable.
Our research spans nonsmooth and stochastic optimization, fixed-point and splitting methods, optimal transport, model compression, and learning for inverse problems, with an emphasis on both theory and scalable computation.
Methods
- Nonsmooth and stochastic optimization
- Operator splitting and fixed-point algorithms
- Structured and parameter-efficient learning
Applications
- Large-scale data science
- Optimal transport and graph learning
- Reliable learning for inverse problems