Information Processing
& Algorithms Laboratory

2024 SPS Best Paper Award!
Featured Article in IEEE Signal Processing Magazine

Algorithm Unrolling: Interpretable, Efficient Deep Learning for Signal and Image Processing by V. Monga, Y. Li and Y. Eldar.
The article can be found here.

Dr. Monga Elevated to IEEE & AAIA Fellow!

Dr. Monga was elevated to IEEE and AAIA Fellow in recognition of his contributions to computationally efficient image analysis and restoration. This honor is the latest in a series of research distinctions, including IEEE best paper awards, a National Science Foundation CAREER Award, and induction into the National Academy of Inventors.
The IEEE Fellow news article can be found here.
The AAIA fellow news article can be found here.

IEEE TBME

GLAPAL-H: Global, Local And Parts Aware Learner for Hydrocephalus Infection Diagnosis in Low-Field MRI

We develop GLAPAL-H, a multi-task deep learning model with global, local, and parts-aware segmentation branches for low-field MRI-based hydrocephalus classification and etiology modeling. A novel three-way regularization strategy — global, local, and parts — enables the model to capture both holistic and fine-grained features while learning soft masks to localize hydrocephalic patterns. GLAPAL-H achieves superior performance over state-of-the-art CT-based and MRI-based methods on both two-class (PIH vs. NPIH) and three-class (PIH vs. NPIH vs. healthy) tasks, with improved interpretability and robustness to data quality. Read More

IEEE TCI

Deep, Convergent, Unrolled Half-Quadratic Splitting For Image Deconvolution

We propose an unrolling technique that breaks the trade-off between retaining algorithm properties while simultaneously enhancing performance. We focus on image deblurring and unrolling the widely-applied Half-Quadratic Splitting (HQS) algorithm. We develop a new parametrization scheme which enforces layer-specific parameters to asymptotically approach certain fixed points. Through extensive experimental studies, we verify that our approach achieves competitive performance with state-of-the-art unrolled layer-specific learning and significantly improves over the traditional HQS algorithm. We further establish convergence of the proposed unrolled network as the number of layers approaches infinity, and characterize its convergence rate. Our experimental verification involves simulations that validate the analytical results as well as comparison with state-of-the-art non-blind deblurring techniques on benchmark datasets. The merits of the proposed convergent unrolled network are established over competing alternatives, especially in the regime of limited training. Read More

NEW! Handbook of Convex Optimization Methods in Imaging Science

This book discusses imaging science and provides tools for solving image processing and computer vision problems using convex optimization methods. Throughout the handbook, the authors introduce topics on the most key aspects of image acquisition and processing that are based on the formulation and solution of novel optimization problems.