| Scientific datasets obtained through numerical simulation (e.g., Computational Fluid Dynamics), geometric modeling (e.g., automobile surface design), and instrumental acquisition (e.g., Computerized Tomography, Magnetic Resonance Imaging, Positron Emission Tomography, Single Photon Emission Computerized Tomography, Ultra-Sound, Confocal Microscopy, LIght Detection And Ranging) are growing at an exponential rate. Without high-performance processing, effective representation, and intuitive rendering of large-scale data, informative comprehension of the underlying phenomenon might not be achieved. Scientific Visualization (SciVis) draws on techniques from computer graphics, image processing, computer vision, human perception, and a wealth of domain-specific knowledge to provide deep insight into complex patterns that otherwise could not be interpreted. As a powerful visual data analysis tool, SciVis has a widespread as well as profound impact on scientific research in many disciplines (e.g, physics, chemistry, medicine, bio-sciences, geo-sciences, and aerospace engineering) and real world applications (e.g., Computer-Aided Diagnosis, tele-collaborative surgery, aircraft design, weather forecast, and geological survey). |
 |
 |
Despite the extensive use in oceanographic-atmospheric modeling, computational fluid dynamics simulation, and electro-magnetic field analysis, flow visualization remains one of the most challenging research topics of SciVis due to the daunting task of conveying the flow direction, particularly in three-dimensional senarios, in a visually understandable way. Thus the major goal is to design novel flow representation schemes to depict directional information in a perceptually effective manner. Important issues include efficiency (computational speed and memory cost), increasingly large datasets and sophisticated demands, time-varying or unsteady flows, complex grids and multivariate (e.g., velocity, temperature, pressure, density, and viscosity) visualization, and feature extraction and tracking. |
 |
 |
Iso-surface (extraction) and (Direct) Volume Rendering are two classical methods for visualizing volume data. Iso-surface techniques such as Marching Cubes, Dividing Cubes, and Marching Tetrahedra work by fitting intermediate geometric primitives (e.g., triangles) to any reconstructed surface prior to a rendering procedure. Volume rendering techniques such as Ray Casting, Splatting, Cell Projection, Shear-Warp, and hardware-based (e.g., Graphical Processing Unit, GPU) 2D / 3D texture mapping eliminate the need for the construction of any intermediate representation, but instead operate directly on voxels (volume elements, the 3D counterpart of "pixels") by employing a light absorption-emission model and a transfer function (i.e., a map / transform used to emphasize or suppress part of the data from the rest) to assign colors and opacities to the voxels that are then sampled and composited or blended (through integral convolution) along the viewing direction. Volume segmentation and registration, automatic but effective transfer function design, and robust feature extraction and tracking are still open problems. |
 |
 |
The advances in protein synthesis, molecular design, and genomic sequencing demonstrate the significance and prospect of bio-technologies. Molecular biology addresses biomolecular structures, properties, functions, relationships, and mechanisms governing bio-chemical processes. Biomolecular visualization helps biologists gain an insightful interpretation of data collected, e.g., using conventional wide-field microscopy or LSCM (Laser Scanning Confocal Microscopy). Low signal-to-noise ratio, insufficient image contrast, longitudinal smearing, intangible amorphous micro-structures, porous surfaces, ragged contours, and lack of a priori knowledge make it a challenging task to visualize confocal microscopic biomolecular data using traditional techniques, such as iso-surface extraction and direct volume rendering, that have been even though well suited for displaying medical CT / MRI data. |
 |
 |
CT (Computerized Tomography) plays an important role in Computer-Aided Diagnosis and healthcare clinics. As a set of rays are cast from a rotating CT scanner to pass through tissues, bones, and whatever organs, the source radiological intensities attenuate to what are measured at the exit by an array of detectors to produce projection data, from which a Filtered Back-Projection method based on Fourier Slice Theorem is typically adopted to reconstruct, in frequency domain, a stack of slice images. Compared to Equi-Angular Fan Beam, Equi-Spatial Fan Beam, Cone Beam, and Single-Slice Helical Scan, Multi-Slice Helical Scan is a more powerful acquisition approach because of the accelerated scanning process (which greatly reduces the patient's discomfort) and the increased temporal-spatial resolution. An ambitious effort in the CT imaging community is to reconstruct a sequence of high-fidelity images of a motion organ (e.g., lung) from multi-slice spiral CT scan. |