Efficient Texture Representation and Sampling Algorithms for Real-time Rendering

Pavlos Mavridis
March 2013
A dissertation submitted to the Department of Informatics of Athens University of Economics & Business


In computer graphics, the goal of photorealistic rendering algorithms is to create convincing images given the description of a scene. However, photorealistic image generation is still an elusive goal for real-time rendering, where the final image should be synthesized in a small fraction of a second. In order to meet this performance requirement, images rendered in real-time often suffer from reduced quality, frequently manifested as low-detailed surfaces, aliasing artifacts and non-realistic illumination.

The goal of this thesis is to improve the quality of real-time rendering through more efficient data representation and sampling algorithms. First, we present more compact representations for textures and frame buffers, reducing the consumed memory bandwidth and allowing more efficient use of the available memory space. Furthermore, we present and investigate a method that improves the surface detail under extreme conditions, such as grazing viewing angles, highly warped texture coordinates, or extreme perspective, using high-quality elliptical texture filtering on the GPU. Finally, we propose the use of a volume-based representation of the scene in order to simulate the diffuse light transport in real-time for completely dynamic scenes. To this end, we present efficient algorithms for the creation and sampling of this volume representation, which is stored as a volume texture on the graphics memory.

Advisor: Prof. Georgios Papaioannou

Teaser Images

The following teaser images, taken from our publications, provide a good visual overview of our work:

Teaser 1. We investigate methods to compute the diffuse indirect illumination using a volume representation of the 3D environment, decoupling the global illumination computations from the complexity of the original dataset. To this end, we present three new efficient voxelization methods and new volume sampling techniques.
Teaser 2. Our sampling method significantrly improves the quality of texture filtering on existing hardware. Left: Our elliptical filtering implementation using conservative quality settings. Right: Hardware annisotropic filtering (NVIDIA GTX460) using the highest possible quality settings in the drivers.
Teaser 3. We introduce a new texture compression scheme that encodes grayscale and color textures at various bitrates, improving the flexibility of existing texture compression formats. The already available decompression hardware is used to make the decoding efficient on existing GPUs and APIs.
Teaser 4. We present a novel frame buffer compression technique that rasterizes a color image using only two frame buffer channels, by interleaving the chrominance components in a checkerboard pattern, as shown in the image. The final frame buffer is reconstructed using an edge-directed demosaicing filter. The compression error, visualized by the heatmap in the inset, is negligible and the filtering is temporally stable.


Demos and Source Code

Demo applications and source code demonstrating our work. These demos generally require a fast PC to run and are tested on Nvidia GPUs.


The work presented in this dissertation appeared in the following scientific journals, international conferences and books:


P. Mavridis, G. Papaioannou, "Texture Compression using Wavelet Decomposition", Computer Graphics Forum (Proceedings of Pacific Graphics 2012), Volume 31, Number 7, September 2012.

P. Mavridis, G. Papaioannou, "The Compact YCoCg Frame Buffer", Journal of Computer Graphics Techniques (JCGT), Volume 1, Number 1, September 2012.


P. Mavridis, G. Papaioannou, "Practical Frame Buffer Compression", in the book GPU Pro 4: Advanced Rendering Techniques, Wolfgang Engel (ed.), AK Peters/CRC Press, 2013.

P. Mavridis, G. Papaioannou, "Practical Elliptical Texture Filtering on the GPU", in the book GPU Pro 3: Advanced Rendering Techniques, Wolfgang Engel (ed.), AK Peters/CRC Press, 2012.

International Conferences

P. Mavridis, G. Papaioannou, "High Quality Elliptical Texture Filtering on GPU", Proceedings of the 2011 ACM SIGGRAPH symposium on Interactive 3D Graphics and Games (i3D 2011).

P. Mavridis, G. Papaioannou, "Texture Compression using Wavelet Decomposition: a Preview", (Poster) Proceedings of the ACM SIGGRAPH Symposium on Interactive 3D Graphics and Games,2012.

A. Gaitatzes, P. Mavridis, G. Papaioannou, "Two Simple Single-pass GPU methods for Multi-channel Surface Voxelization of Dynamic Scenes", (short paper) Pacific Graphics 2011.

P. Mavridis, G. Papaioannou, "Global Illumination Using Imperfect Volumes", (short paper) Proceedings of the International Conference on Computer Graphics Theory and Applications (GRAPP), 2011

P. Mavridis, A. Gaitatzes, G. Papaioannou, "Volume-based diffuse global illumination". In Proc. CGVCVIP '10: Proceedings of Computer Graphics, Visualization, Computer Vision and Image Processing 2010.

A. Gaitatzes, P. Mavridis, G Papaioannou, "Interactive Volume-based Indirect Illumination of Dynamic Scenes", Proc. 3IA '10: Proceedings of the 2010 International Conference on Computer Graphics and Artificial Intelligence, May 2010.