Over the last decade Visual Computing has established itself as a
core discipline within computer science and information
technology. Its research scope covers the design of computational and
mathematical methods and interactive systems for geometric and
physics-based modeling, image acquisition and synthesis, scientic
visualization, as well as virtual reality. As a result, visual
computing research is providing enabling technology for a broad
spectrum of applications in all of CRS4 thematic areas.
The Visual Computing Group has been founded in 1996, and has
developed remarkable skills in visual computing technology,
gaining a national and international visibility.
Current projects range from desktop 3D
tools to immersive virtual environments with haptic
feedback. Representative examples of current active research domains
are the following:
Rendering and streaming of terrains and urban environments. We
develop multi-resolution out-of-core techniques for rapid high-quality
visualization of textured digital terrain models and high resolution
urban environments. For terrain rendering, we developed the Batched
Dynamic Adaptive Meshes Framework (BDAM). The framework introduced one
of the first methods for rapidly generating seamless variable
resolution surfaces by assembling precomputed patches (BDAM -
Eurographics 2003). It exploits GPU programmability for ensuring
submetric accuracy on full planet visualization (P-BDAM - IEEE
Visualization 2003). Coupled with a two-stage wavelet based near
lossless compression scheme, the method supports data compression with
maximum error metric (C-BDAM - Eurographics 2006). The efficiency of
the approach has been demonstrated on a number of local
and global test cases, and is at the core of successfully
deployed internet geoviewers. For urban model rendering and
streaming, we introduced a GPU-friendly technique that
efficiently exploits the highly structured nature of urban
environments to ensure rendering quality and interactive performance
of city exploration tasks (BlockMap - Eurographics 2007). Central to
our approach is a novel discrete representation for the efficient encoding and
rendering of a small set of textured buildings far from the viewer.
Work done in collaboration with ISTI-CNR.
Processing, distribution, and rendering of massive dense 3D meshes and point clouds.
We look at techniques for supporting
inspection of surface models characterized by a high sample density,
such as those generated by laser scanning. Our main result is the
introduction of a coarse grained multiresolution model based on
hierarchical volumetric decomposition, that lead to the first GPU
bound high quality technique for large scale meshes (2004). Introduced
optimized representation that exploits the properties of a conformal
hierarchy of tetrahedra (Adaptive TetraPuzzles, SIGGRAPH 2004) and a
general framework based on an extension of the Multi-Triangulation
(GPU-MT, IEEE Visualization 2005). We also introduced the first
coarse-grained multiresolution point hierarchy (LPC, Computers and
Graphics 2004). Demonstrated on a number of test cases, including
all Digital Michelangelo models. Work done in collaboration with
Processing, distribution, and rendering of huge complex 3D models.
We focus on methods able to support very large arbitrary surface models
with high topological genus, highly variable depth complexity, fine geometric detail, "Bad"
tessellations. Models of this kind arise from numerical simulation and
computer aided design. Our main result is the introduction of a
volumetric method based on multi-scale modeling of appearance rather
than geometry with tight integration of visibility and LOD
construction (Far Voxels, SIGGRAPH 2005). The method exploits GPU
programmability for accelerated rendering and has been demonstrated on
a number of test cases, ranging from laser scans, to isosurfaces, to
extremely large CAD models, including the full Boeing 777 model. Fully
interactive performance even for large windows on current single
Massive volume rendering.
We focus on methods able to render models of potentially unlimited
size on current GPU platform. Our methods are based on adaptive
out-of-core multiresolution techniques with visibility feedback
realized within a single-pass GPU raycasting framework.
Specialized techniques have been created both within an OpenGL
shading framework (Visual Computer 2008) and NVIDIA CUDA (Visual
Computer 2009). Illustrative techniques have been realized
to increase volume understanding. Fully interactive performance
has been demonstrated on datasets of many GVoxels.
Interactive visualization on novel light field displays.
We focus on developing efficient techniques for harnessing the power of novel 3D display
design. We work with Holografika (Hungary), that develops a 3D
display combining a specially arranged array of projectors and a holographic screen. By
properly controlling image generation, rendered objects appear
floating in space to multiple naked eye viewers. Sustaining
interactive rates is a challenging tasks, since hundreds of views per
frame have to be generated. We develop specialized surface and volume
rendering techniques for both single processor and network parallel
rendering. By dynamically adapting resolution to spatial display
capabilities, we have demonstrated the capability to
sustain interaction for huge models on a 50Mpixel display (SIGGRAPH
2006 Etech). We have subsequently designed and produced specialized
interactive rendering systems both for meshes and volumetric data
(Eurographics 2008, Visual Computer 2009, 2010).
Real-time surgical simulation with visual and haptic feedback.
We develop enabling technology to support
surgical training through simulation. Our results so far include a
multiresolution volumetric model for simulating bone burring,
simplified models for contrast agent transport in human vessels,
real-time techniques for phacoemulsification simulation. Some of our
simulation and rendering modules have been integrated in industrial
systems for surgical training.
Check our projects, publications, and
for more information on our activities.