Deep unfolding network for image restoration by quantum interactive patches
| Quand ? |
Le 10/05/2022, de 13:00 à 14:00 |
|---|---|
| Participants |
Adrian Basarab |
| Ajouter un événement au calendrier |
vCal
iCal
|
Title : Deep unfolding network for image restoration by quantum interactive patches
Asbtract : Decomposition of digital signals and images into particular basis or dictionaries is a very useful tool in number of applications. Such a decomposition is commonly obtained using fixed transforms (e.g., Fourier, DCT or wavelet) or data-driven dictionaries learned from example databases or from the signal or image itself (e.g., by exploiting the redundancy within patches extracted from one or several images). This talk will present a new idea of constructing such a signal or image-dependent bases inspired by quantum mechanics tools, i.e., by considering the signal or image as a potential in the discretized Schrödinger equation. This quantum signal or image processing method based on the theory of single particle quantum systems is further generalized to patch-wise image representation using the concepts of quantum many-body interaction. The similarity between two image patches is introduced in the formalism through a term akin to interaction terms in quantum mechanics. Finally, a deep unfloding deep learning network is designed to integrate these concepts and allow more flexibility in the choice of the hyperparameters.The potential of the proposed framework is illustrated through denoising, deconvolution and super-resolution results in presence of Gaussian, Poisson, and speckle noise.
More information : https://www.irit.fr/~Adrian.Basarab/ ; working on computational medical imaging at CREATIS (Lyon)