Roman Jacome

[Roman Jacome @alejandro033099]

I am glad to be selected as a Outstanding Reviewer in #CVPR2026

[#CVPR2026 @CVPR]

We are grateful to all of the 17,491 reviewers who helped make #CVPR2026 possible. We are especially pleased to recognize the following Outstanding Reviewers, whose high-quality reviews (as judged by their Area Chairs) placed them among the top 5% of reviewers.

[Roman Jacome @alejandro033099]

We incorporate GSNR in different solvers, such as PnP, DM, and DIP, with different sensing operators, showing consistent improvements in performance and convergence.

[Roman Jacome @alejandro033099]

Glad to share that our recent work on inverse problems was accepted to #CVPR2026

📜GSNR: Graph Smooth Null-Space Representation for Inverse Problems.

@GualdronHurtado Rafael S. Suarez and Prof @henarfu

https://t.co/vs2Ryzmebe

[Roman Jacome @alejandro033099]

i) Coverage: With a few eigenmodes of the null-restricted Laplacian, we cover most of the null-space energy.

ii) Minimax optimal: the GSNR basis is worst-case optimal over a graph-energy ellipsoid

iii) Predicatability: GSNR has a higher per-mode recoverability from measurements

[Roman Jacome @alejandro033099]

We propose a new regularization strategy that learns the low-dimensional null-space structure and can be integrated into any pipeline—PnP, diffusion, deep image prior, or unrolling—providing orthogonal guidance to data fidelity

📜https://t.co/LXwxbkxClt📜

[Roman Jacome @alejandro033099]

Delighted to share that our work on inverse problems in imaging was accepted at #NeurIPS 2025.

NPN: Non-Linear Projections of the Null-Space for Imaging Inverse Problems
Roman Jácome*, @GualdronHurtado , @leonsuarez_24, Henry Arguello
(*equal contribution).

[Roman Jacome @alejandro033099]

Accepted to IEEE TCI. Using KD to design better computational imaging systems.
We show how a teacher–student setup guides the design of physical operators, improving recovery quality and efficiency across inverse problems. @leonsuarez_24 and Prof @henarfu https://t.co/qHp15iEsMM

[Roman Jacome @alejandro033099]

I'm glad to share our recent work, "D2GP: Deep Distillation Gradient Preconditioning for imaging inverse problems". A principled teacher–student framework learns a nonlinear preconditioner, boosting convergence and reconstruction. Accepted to CAMSAP 2025 https://t.co/Pq3lZWgeGe

[Roman Jacome @alejandro033099]

I want to share our new paper on Diffusion Consensus Equilibrium for sparse-view CT. Balances data consistency + diffusion priors through a two-agent consensus scheme → more robust reconstructions under severe undersampling. Accepted to CAMSAP 2025 https://t.co/B4TiM4LKu9

[Roman Jacome @alejandro033099]

I want to share our new paper. A homotopy-based training strategy for unrolled networks in imaging inverse problems. Smoothly transitioning from well-posed to ill-posed settings leads to better performance & generalization. Accepted to CAMSAP 2025 https://t.co/Orxbcb0KMM

[Roman Jacome @alejandro033099]

In this work, we proposed a spatial-spectral shift variant multi-shot spectral imaging system by including a double DOE to codify the phase of the incident wavefront. The DOE structure is jointly optimized with an unrolling network for reconstruction.
https://t.co/gqvMCcduGZ

[Roman Jacome @alejandro033099]

📜Glad to share our recent IEEE JSTSP paper "DoDo: Double DOE Optical System for Multishot Spectral Imaging" in collaboration with @SergioUrrea99 (UIS, CO), Prof @salmanasif (UCR, USA), Prof @henarfu (UIS, CO), and Prof Hans Garcia(UIS, CO).📜

[Roman Jacome @alejandro033099]

Glad to share our conference paper at #ICASSP2024 "Multi-Antenna ISAC Receiver with n-Tuple Blind Deconvolution" This is a joint work with Edwin Vargas, Kumar Vijay Mishra, Brian M Sadler, and
@henarfu
https://t.co/rbDLrbm2vl

[Roman Jacome @alejandro033099]

We proposed a set of regularization functions that enables proper optical design by exploiting statistical properties of the coded measurement distribution during the training stage. We outperform state-of-the-art methods in several optical systems and various computational tasks

[Roman Jacome @alejandro033099]

I am glad to share our new work "Middle output regularized end-to-end optimization for computational imaging" on deep optical coding design published in Optica. This was a joint work with Pablo Andres Gomez Toloza and Prof Henry Arguello.
https://t.co/0h05DVwnTJ

[Roman Jacome @alejandro033099]

Here we solved a dual blind deconvolution problem for an ISAC evolving in time leading to a dynamic setting. We proposed an efficient algorithm based on expectation maximization over a factor graph structure.

[Roman Jacome @alejandro033099]

I am glad to share our new work on integrated sensing and communications (ISAC) that will be presented at IEEE CAMSAP 2023.
This is a joint work with Edwin Vargas, Kumar Vijay Mishra, Brian M Sadler and @henarfu

https://t.co/wGftQryUsD