High fidelity walkthroughs in archaeology sites

Abstract

Fast and affordable computing systems currently support walkthroughs into virtual reconstructed sites, with fast frame rate generation of synthetic images. However, archaeologists still complain about the lack of realism in these interactive tours, mainly due to the false ambient illumination. Accurate visualizations require physically based global illumination models to render the scenes, which are computationally too demanding. Faster systems and novel rendering techniques are required: current clusters provide a feasible and affordable path towards these goals, and we developed a framework to support smooth virtual walkthroughs, using progressive rendering to converge to high fidelity images whenever computing power surplus is available. This framework exploits spatial and temporal coherence among successive frames, serving multiple clients that share and interact with the same virtual model, while maintaining each its own view of the model. It is based on a three-tier architecture: the outer layer embodies light-weight visualization clients, which perform all the user interactions and display the final images using the available graphics hardware; the inner layer is a parallel version of a physically based ray tracer running on a cluster of off-the-shelf PCs; in the middle layer lies the shading management agent (SMA), which monitors the clients' states, supplies each with properly shaded 3D points, maintains a cache of previously rendered geometry and requests relevant shading samples to the parallel renderer, whenever required. A prototype of a high fidelity walkthrough in the archaeologic virtual model of the roman town of Bracara Augusta was developed, and the current evaluation tests aimed to measure the performance improvements due to the use of SMA caches and associated parallel rendering capabilities. Preliminary results show that interactive frame rates are sustainable and the system is highly responsive.