Volume graphics shaders for GPU

• Main Author: Muhammad Mobeen Movania
• Other Authors: Lin Feng
• Format: Theses and Dissertations
• Language: English
• Published: 2012
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Computer graphics; DRNTU::Engineering::Computer science and engineering::Computing methodologies::Simulation and modeling
• Online Access: https://hdl.handle.net/10356/50773
• Institution: Nanyang Technological University

Volume graphics attracts our research interest. On one hand, the challenging tasks in cyberspace applications are to model and render the objects and phenomena with complex properties such as the nonuniform and nonrigid materials in volumetric datasets. On the other hand, graphics algorithms have been developed mainly for rigid object representation, affine transformation and offline rendering. This limits their applications in time-constraint yet complex systems, for example, image-guided surgery and therapy systems. In response, this thesis presents a comprehensive study on modeling and rendering of volumetric graphical objects and the acceleration technologies for the new-generation Graphics Processing Unit (GPU). The research focus is on the novel hardware accelerated solutions by programming shaders on the GPU, including the vertex, tessellation, geometry and fragment shaders. First, there is a strong demand for integrating the volume visualization and deformation processes in real-time systems. We are interested in an efficient method that can utilize the new features of the modern GPU. One such feature is transform feedback which is a special mode in which the GPU feedbacks the vertices in its own clock cycles. While this mode was initially used for dynamic tessellation and Level-Of-Detail (LOD) rendering, we have exploited this mode for a deformation pipeline implemented entirely on the GPU using transform feedback. Our experimental results reveal that the proposed pipeline outperforms other GPU implementation schemes. Moreover, we can couple such a deformation pipeline with the visualization pipeline seamlessly reducing the amount of data transfer out of the GPU core. Previous approaches suffer from an unbalanced utilization of the graphics pipeline; that is, they are either vertex shader bound or fragment shader bound. On the contrary, since our pipeline uses the vertex shader stage for deformation and uses the fragment shader stage for volume rendering, it makes a balanced utilization of the programmable graphics pipeline. We apply the proposed pipeline first in a mesh based approach for large deformation and then extend it to a meshless approach for relatively small and accurately controlled deformation.