Project start: 1.10.2025.

Three-dimensional (3D) scenes have become increasingly vital across various fields, including film, medicine, and virtual reality, due to their immersive and realistic visualizations. These scenes are created through 3D modeling and rendering, transforming virtual models into images. Advanced textures, lighting effects, and interactive capabilities simulate real-world environments, objects, and people, providing an engaging experience for users. However, creating high-quality 3D scenes poses challenges such as high processing and storage requirements, complex geometries, and artifact suppression.

This research aims to enhance 3D scene realism by integrating light fields (LFs), which capture spatial and angular scene information from multiple viewpoints, enabling 3D scene reconstruction. Leveraging the advantages of compressive sensing (CS) in signal and image processing, the primary objective is to develop and implement novel CS techniques for LF rendering in 3D scenes. The focus will be on reconstructing LFs from limited data and cameras, allowing high-quality rendering with reduced data. It is anticipated that this will lead to reductions in data storage needs and improvements in rendering efficiency. To further enhance rendering speed and quality, deep learning methods will be analyzed and implemented. Deep learning will facilitate real-time rendering and interactive applications, providing a virtual reality experience.

The practical aspect of this project involves acquiring specialized display and devices for visualizing 3D scenes. This equipment will not only aid in visualizing research results but also contribute to establishing a new research direction at the University of Rijeka, improving the teaching content, promoting science, and fostering collaboration with Tampere University, Finland. The collaboration will leverage expertise in 3D scene reconstruction and visualization, facilitating joint research projects and publications.