Sparse Reconstruction

Sparse reconstruction aims to recover high-dimensional signals from limited, often noisy, measurements by leveraging the inherent sparsity (few non-zero components) of the underlying data. Current research focuses on developing efficient algorithms, including those based on state-space models, orthogonal matching pursuit, and iterative methods like gradient projection and ADMM, often integrated with deep learning architectures such as convolutional neural networks and specialized networks designed for specific signal types (e.g., handling aliasing artifacts in MRI). These advancements improve the speed and accuracy of reconstruction in various applications, including medical imaging (MRI, ODT), radar imaging, and signal processing, leading to better image quality and reduced data acquisition times.