University of Oulu

Scene relighting and editing for improved object insertion

Saved in:
Author: Gabr, Shrief1
Organizations: 1University of Oulu, Faculty of Information Technology and Electrical Engineering, Computer Science
Format: ebook
Version: published version
Access: open
Online Access: PDF Full Text (PDF, 12.5 MB)
Pages: 54
Persistent link:
Language: English
Published: Oulu : S. Gabr, 2023
Publish Date: 2023-06-16
Thesis type: Master's thesis (tech)
Tutor: Heikkilä, Janne
Nguyen, Phong
Reviewer: Heikkilä, Janne
Mustaniemi, Janne


The goal of this thesis is to develop a scene relighting and object insertion pipeline using Neural Radiance Fields (NeRF) to incorporate one or more objects into an outdoor environment scene. The output is a 3D mesh that embodies decomposed bidirectional reflectance distribution function (BRDF) characteristics, which interact with varying light source positions and strengths. To achieve this objective, the thesis is divided into two sub-tasks.

The first sub-task involves extracting visual information about the outdoor environment from a sparse set of corresponding images. A neural representation is constructed, providing a comprehensive understanding of the constituent elements, such as materials, geometry, illumination, and shadows. The second sub-task involves generating a neural representation of the inserted object using either real-world images or synthetic data.

To accomplish these objectives, the thesis draws on existing literature in computer vision and computer graphics. Different approaches are assessed to identify their advantages and disadvantages, with detailed descriptions of the chosen techniques provided, highlighting their functioning to produce the ultimate outcome.

Overall, this thesis aims to provide a framework for compositing and relighting that is grounded in NeRF and allows for the seamless integration of objects into outdoor environments. The outcome of this work has potential applications in various domains, such as visual effects, gaming, and virtual reality.

see all

Copyright information: © Shrief Gabr, 2023. Except otherwise noted, the reuse of this document is authorised under a Creative Commons Attribution 4.0 International (CC-BY 4.0) licence ( This means that reuse is allowed provided appropriate credit is given and any changes are indicated. For any use or reproduction of elements that are not owned by the author(s), permission may need to be directly from the respective right holders.