Technical University of Crete (TUC)
Implicit and foveated techniques for real-time computer graphics rendering
Abstract
dc:descriptionReal-time rendering is a cornerstone of modern interactive applications, such as virtual reality, gaming, and simulations, where high visual fidelity and low latency are critical to user immersion. However, achieving these goals remains a significant challenge due to the computational complexity of rendering techniques and the limitations of existing methods. Traditional approaches often struggle to balance performance and quality, particularly in scenarios requiring dynamic adaptation to user attention or the rendering of complex geometries. These challenges are especially pronounced in foveated rendering, which aims to optimize resource allocation by prioritizing high-quality rendering in the user's focal region, and in sphere tracing of Signed Distance Functions (SDFs), a technique used to render intricate implicit surfaces. Current methods frequently suffer from visual artifacts, inefficient resource utilization, and an inability to adapt to real-time demands, highlighting the need for innovative solutions that bridge the gap between computational efficiency and perceptual quality. This doctoral dissertation presents innovative approaches designed to enhance the efficiency of real-time rendering techniques, focusing on foveated rendering and sphere tracing of Signed Distance Functions (SDFs). The research addresses the limitations of existing rendering methods that rely heavily on low-level image features and often suffer from artifacts, when optimizing computational resource allocation. First, we present an emulated foveated path tracing framework that establishes thresholds for imperceptible image manipulations based on user gaze. Our perceptual studies yield varying eccentricity thresholds for foveated performance, highlighting the impact of experimental methodologies on user sensitivity to visual changes. The results indicate potential computational complexity reductions of at least $2\times$ - $3\times$ in path tracing performance through foveated rendering methods. Second, we introduce the Foveated Inverted Pyramid Rendering (FIPR) system, which optimizes rendering quality based on user gaze. By employing a multi-scale inverted pyramid structure, we leverage low-resolution renderings to incrementally refine ray distances, dramatically reducing the overall ray step count and enabling efficient rendering of complex scenes in virtual reality. Our method supports $16\times$ super-sample anti-aliasing, while maintaining imperceptible image quality transitions, even in peripheral vision. Lastly, we develop a novel rapid rendering pipeline for sphere tracing SDFs that significantly reduces the overall ray step count using ultra-low resolution renderings, while minimizing artifacts. By employing a single low-resolution buffer and scaling SDFs within this buffer, our method ensures the visibility of small features while enabling earlier ray termination for high-cost surface edges. This approach yields a substantial performance improvement, achieving speedups exceeding $3\times$ compared to traditional methods. Collectively, these contributions advance the field of real-time graphics rendering, providing valuable insights into optimizing rendering techniques, while enhancing the user experience in interactive applications, such as gaming and simulations.
Degree
thesis:*- Grantor dc:publisher
- Technical University of Crete (TUC)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Polychronakis, Andreas
- Πολυχρονάκης, Ανδρέας
Subjects
dc:subject × 11- Μη πολυγωνικές επιφάνειες
- Implicit surface rendering
- Real time rendering
- Sphere tracing
- Foveated rendering
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Γραφικά υπολογιστή και Σχεδιασμός με χρήση υπολογιστή
- Computer and Information Sciences
- Natural Sciences
- Computer Graphics and Computer-Aided Design
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/58598