Reconstruction of real-world scenes from a set of multiple images is a topic in computer vision and 3D computer graphics with many interesting applications. Attempts have been made to real-time reconstruction on PC cluster systems. While these provide enough performance, they are expensive and less flexible. Approaches that use a GPU hardware-acceleration on single workstations achieve real-time framerates for novel-view synthesis, but do not provide an explicit volumetric representation. This work shows our efforts in developing a GPU hardware-accelerated framework for providing a photo-consistent reconstruction of a dynamic 3D scene. High performance is achieved by employing a shape from silhouette technique in advance. Since the entire processing is done on a single PC, the framework can be applied in mobile environments, enabling a wide range of further applications. We explain our approach using programmable vertex and fragment processors and compare it to highly optimized CPU implementations. We show that the new approach can outperform the latter by more than one magnitude and give an outlook for interesting future enhancements.

Caustics are patterns of light focused by reflective or refractive objects. Because of their visually fascinating patterns, several methods have been developed to render caustics. We propose a method for the quick rendering of caustics formed by refracted and converged light through transparent objects. First, in the preprocess, we calculate sampling rays incident on each vertex of the object, and trace the rays until they leave the object taking refraction into account. The position and direction of each ray that finally transmits the transparent object are obtained and stored in a lookup table. Next, in the rendering process, when the object is illuminated, the positions and directions of the rays leaving the object are calculated using the lookup table. This makes it possible to render refractive caustics due to transparent objects at interactive frame rates, allowing us to change the light position and direction, and translate and rotate the object.

We present a scalable parallel rasterizer based on our interleaved scanline rasterization. The sorting overhead of a conventional scanline-based parallel rendering approach has been studied and removed by implementing a scanline assignment hardware. All advantages of the scanline-based parallel rendering are kept such that a good scalability and a small memory usage are achieved. Our architecture is evaluated precisely by a discrete event-based simulation, and the rendering performance and utilization are shown for a various number of rasterizers. The simulation results show more than 8 Mtriangles/s of performance with 64 rasterization engines running at 10 MHz.