Omar Elezabi

I'm a PhD Researcher in Computer Science working on Computer Vision problems at the University of Würzburg, advised by Prof. Radu Timofte.

For MSc I was awarded a full scholarship from Erasmus Mundus Joint Masters. I obtained my MSc degree in Computational Colour and Spectral Imaging from NTNU (Norway) and Jean Monnet University (France).
Prior I resived a Bacholar in Computer Engineering from Mansoura University

CV  /  Mail  /  Google Scholar  /  GitHub  /  LinkedIn

News

Research

My current research interests include deep learning, low-level computer vision, generative image restoration/enhancemnt, and computational photography.
My current research focus is making professional quality photography accessible for normal users
Representative papers are highlighted -- check my Google Scholar to see the complete list.

We propose VDC, a method to adapt text-conditioned diffucsion models for image editing and restoration tasks using only one visual example. Our method addresses the issue of text-visual alignment in diffusion models, allowing data-efficient and fast adaptation.

We propose INRetouch, a context-aware implicit neural representation for photography retouching. Our method uses neural representations to learn and apply photo retouching styles.

We poropose a new paradiam to solve the misalginemnt issue in Real World SR datasets. Our pipleine create a mimicked LR from the HR image creating a full aligned training data producing a sharper output without artifacts

First, we propose a novel module that can be integrated into any neural ISP to capture the global context information from the full RAW images. Second, we propose an efficient and simple neural ISP that utilizes our proposed module.

We investigate observer metamerism (OM) in color displays and proposes a quantitative assessment of the OM distribution across the chromaticity diagram.

This work proposes a texture classification method applied directly to the raw multispectral image. Additionally, we investigate the role of integrating time and illumination on the performance.

A deep learning based approach is proposed to detect the driver’s actions and eliminate the driver’s distraction as a packed solution. (Bachelor's Thesis)

This paper introduces an automatic roadway safety features detection approach using an oriented (or rotated) object detection model.

Academic Service

Teaching: Image Processing and Computational Photography (IPCP), Computer Vision (CV)

Reviewer: CVPR, ICCV, ECCV.

Workshops: NTIRE CVPR 2026  /  NTIRE CVPR 2023

Design and source code from Jon Barron's website.