From Quantum Wave Functions to Energy-Efficient Deep Learning: A Quantum-Inspired Journey in Image Denoising
About this Topic:
Image denoising requires representations that effectively separate noise from meaningful information while preserving fine structures, textures, and diagnostically relevant features. This work presents a unified quantum-inspired framework that progresses from adaptive analytical representations to computationally efficient deep learning. Signals and images are first modeled as potentials within the Schrödinger equation, whose wave functions provide data-dependent orthonormal bases for representing spatial and spectral information. This formulation enables robust restoration under correlated or uncorrelated noise. The framework is subsequently extended using concepts from quantum many-body theory, in which interactions among neighboring image patches encode structural similarity and facilitate the preservation of local details across natural and medical images. These physical principles are further translated into an orthonormal convolutional neural network that combines norm-preserving channel transformations with decorrelated feature representations. The resulting architecture maintains stable information flow, reduces feature redundancy, and preserves anatomical fidelity while substantially lowering computational and energy requirements. Collectively, these developments establish a coherent connection between quantum physical models and modern deep learning, providing an interpretable and sustainable foundation for signal restoration and medical image denoising.
About the Presenter:
Sayantan Dutta received the B.Sc. degree in mathematics from the University of Burdwan, Burdwan, India, the M.Sc. degree in applied mathematics from Visva-Bharati University, Santiniketan, India, the M.S. degree in fundamental physics from the University of Tours, Tours, France and the Ph.D. degree in computer science from Université Toulouse III–Paul Sabatier, Toulouse, France in 2016, 2018, 2019 and 2023 respectively. . His doctoral research focused on image-restoration methods inspired by concepts from quantum mechanics.
He is currently a Senior AI Scientist developing artificial intelligence and computational-imaging methods for medical imaging at the Science and Technology Organization, GE HealthCare, Bengaluru, India since 2024. From 2023 to 2024, he was a Postdoctoral Associate with the Department of Radiology, Weill Cornell Medicine, New York, NY, USA, where he worked on quantitative ultrasound and acoustic-microscopy imaging. His research interests include quantum computing, quantum-inspired image processing, computational imaging, medical image restoration, deep learning, and compute- and energy-efficient artificial intelligence.
Want to learn more about upcoming events & webinars?
Visit the events section of the Signal Processing website to see all upcoming lectures, workshops, webinars, and more.