Reading Group

The PICSA Reading Group is a series of research talks and paper discussions organized within the PICSA group at GIPSA-lab. It brings together researchers, students, and visiting colleagues to share recent advances in computational imaging, inverse problems, spectral imaging, multimodal fusion, and machine learning, exchange ideas, and explore new collaborations.

Everyone interested is welcome to attend. MIRA hosts this page for the group.

Upcoming Sessions

No upcoming sessions announced yet.

Past Sessions

2026

Brayan Monroy (UIS, Colombia)

Self-supervised Learning in Image Denoising with GR2R and L2R

· 13:00 · Room B153, GIPSA-lab

Presenter: Brayan Monroy (UIS, Colombia)

Abstract

Learning to denoise images without clean reference data is essential in scientific imaging, where ground truth is often unavailable. This talk presents two complementary frameworks that train denoisers directly from noisy observations: Generalized Recorrupted-to-Recorrupted (GR2R) and Learning to Recorrupt (L2R).

GR2R extends recorruption-based self-supervision beyond Gaussian noise to broader additive noise models and the natural exponential family, including Poisson and Gamma distributions. By constructing training pairs from individual noisy observations, it yields objectives equivalent in expectation to supervised losses, up to model-independent constants, under appropriate conditions. Its connection to Stein’s unbiased risk estimation provides a theoretical foundation without requiring explicit divergence computation.

L2R addresses a central limitation of this approach: the need for prior knowledge of the noise distribution. It introduces a learnable recorruptor jointly optimized with the denoiser through a min–max objective, adapting auxiliary noise statistics directly from noisy data. Experiments demonstrate its effectiveness for non-Gaussian, heavy-tailed, and spatially correlated noise, alongside competitive performance on real imaging benchmarks with signal-dependent noise.

The talk will discuss the theoretical principles, practical implementation, and experimental evidence underlying this progression from known to learned recorruption.

About the Speaker

Brayan Monroy is a doctoral researcher in Computer Science at Universidad Industrial de Santander (UIS), Colombia, and a visiting researcher at GIPSA-lab.

His research combines mathematical optimization and deep learning to address imaging inverse problems, with a particular focus on self-supervised methods for linear and nonlinear forward operators. His work includes self-supervised image denoising beyond Gaussian noise, compressive spectral imaging, and deep-learning approaches for chronic wound assessment. He holds B.Sc. and M.Sc. degrees in Systems Engineering from UIS and contributes to DeepInverse, an open-source Python library for solving imaging inverse problems using deep learning.

Exequiel Oliva (PUCV, Chile)

Multispectral Extended Depth-of-Field Imaging via Stochastic Wavefront Optimization

· 13:00 · Room B153, GIPSA-lab

Presenter: Exequiel Oliva (PUCV, Chile)

Abstract

Extended depth-of-field (EDoF) is a desirable attribute for imaging systems where all features in the scene are in focus despite their relative distance. Traditional imaging systems can achieve EDoF by reducing the aperture size at the expense of signal-to-noise ratio, particularly relevant in spectral imaging systems where incoming light is further divided. By designing and integrating diffractive optical elements (DOEs) placed at the aperture plane of the imaging system, wavefront coding has enabled EDoF while maintaining a larger aperture size at the expense of post-processing. Nevertheless, chromatic aberrations may appear and can often be confused with defocus, jeopardizing reconstruction fidelity.

This work presents a novel design approach for a multispectral-aware DOE for EDoF. By considering and modeling a refractive-diffractive optical setup, the proposed system uses a stochastic optimization framework to optimize DOE patterns, simultaneously preserving spectral fidelity and extending the depth of field. The optimization process exploits the covariance matrix adaptation evolution strategy (CMA-ES), efficiently exploring complex, high-dimensional phase configurations without requiring explicit gradient information.

The optimized DOE is evaluated in a simulated imaging pipeline, where the EDoF multispectral datacube is deblurred using Richardson-Lucy deconvolution. Both qualitative and quantitative results demonstrate that the proposed DOE significantly improves depth invariance and spectral fidelity compared to conventional and state-of-the-art DOE designs, offering a cost-effective solution for real-world multispectral EDoF applications.

About the Speaker

Exequiel Oliva is a Ph.D. student at Pontificia Universidad Católica de Valparaíso (PUCV), Chile, visiting GIPSA-lab.

Exequiel Oliva received his B.Sc. degree in 2025 and Engineering Diploma in Electronics Engineering in 2026 from Pontificia Universidad Católica de Valparaíso (PUCV), Chile, where he is currently pursuing his Ph.D. in Electrical Engineering. His research interests include computational imaging, inverse problems, computational and diffractive optics, wavefront coding, multispectral and hyperspectral imaging, and deep learning for image reconstruction. His current research focuses on the joint design of optical elements and computational reconstruction algorithms, particularly using diffractive optical elements for spectral and spatial imaging applications. He is a Student Member of IEEE and OPTICA.

Organizers