IEEE Computer Graphics and Applications

• PrePrint: Fast Rendering of Diffusion Curves with Triangles
  Diffusion curves are a new kind of primitive in vector graphics, capable of representing smooth color transitions among boundaries. Their rendering requires solving Poisson's equation, and many previous efforts relied on traditional solvers which commonly require GPU acceleration to achieve real-time rasterization. This obviously restricts deployment on the internet, e.g. as Rich Internet Applications (RIAs), in which various computing environments are involved. Inspired by the observation that diffusion effects have very similar appearance to locally defined interpolation with particular orientation and magnitude, we propose a mesh-based approach combined with mean value coordinates (MVC) interpolant to efficiently render diffusion curve images on CPU. A visibility algorithm is employed to efficiently find and sort neighboring curve nodes for each vertex, and then the colors of vertices are assigned according to MVC interpolation with neighboring curve nodes. Our experiments show comparable rendering results to traditional solvers, while our method is computationally more efficient and runs much faster on a CPU.
  
  
  
  
• PrePrint: Practical Noise Reduction for Progressive Stochastic Ray Tracing with Perceptual Control
  We present a method to reduce noise in stochastic ray tracing that is especially tailored to interactive progressive rendering. High-variance light paths are accumulated in a separate buffer, which is filtered by a high-quality edge preserving filter. Then a combination of the noisy unfiltered samples and the less noisy (but biased) filtered samples is added to the low-variance samples in order to form the final image. A novel per-pixel blending operator combines both contributions in a way that respects a user-defined threshold on perceived noise. Our method is able to provide fast, reliable previews, even in the presence of complex features like specular surfaces and high-frequency textures. At the same time it is consistent in the sense that the bias due to filtering vanishes in the limit.
  
  
  
  
• IEEE Computer Graphics and Applications - January/February 2012 (Vol. 32, No. 1)
  IEEE Computer Graphics and Applications
  
  
  
  
• PrePrint: Verifying and Modeling the Polarising Reflectance of Real-World Metallic Surfaces
  Computer Graphics has a number of analytical BRDF models at its disposal, some of which are widely considered to be at least reasonably physically plausible. Significant effort has also been invested in measuring the actual reflectance of a wide range of material surfaces. While such measurements have been used to fit ad hoc combinations of BRDF models to the observed reflectance behaviour, we are not aware of a systematic effort to verify the predictions of basic analytical BRDF models based on measurements of real-world samples. In this paper we propose to use ellipsometry as a way to verify both the actual polarising effect as well as the overall reflectance behaviour of metallic surfaces.
  
  
  
  
• PrePrint: Motion Transplantation Techniques: a Survey
  During the last decade, several techniques have been developed for transplanting motions. Here, a partial auxiliary motion, possibly defined for a small set of degrees of freedom, is transplanted on a base motion. Motion transplantation improves the expressiveness of motion databases and allows more control in interactive applications as various body parts can be synthesized separately. However, the auxiliary motion needs to properly aligned with the base motion, both temporally and spatially. In this paper, we provide an overview of example-based motion transplantation techniques and elaborate on how they determine a spatial and temporal alignment between the auxiliary and base motion. Furthermore, we will elaborate on hybrid techniques that are able to transplant resulting motions from procedural or physics-based techniques.
  
  
  
  
• PrePrint: Salts Decay Aging of Buildings
  Human-made constructions appearance is very dependent on various aging phenomena, resulting in complex patterns that vary with time and environment. Handling such complexity is important in computer graphics in order to improve the realism of virtual scenes. Artists often do this difficult work by hand. Their results are often very good, at the price of two main problems that need to be solved. First, designing specific aging patterns is a tedious task. Also, a large amount of bibliographical work, in various scientific fields, is needed to put these patterns in the correct locations on damaged objects, to obtain plausible results. Another way to tackle this problem is to provide new aging algorithms. In this paper, we address salt-based aging, which leads to very rich and common changes in appearance, and plays an important part in the realism of scenes featuring stone structures and monuments. Our model is physically-inspired: it is based on physical behaviors and principles. It leads to plausible results, replacing the simulation of complex physical formulations by ad-hoc algorithms. An important purpose is to help designers locate aging patterns on affected objects.
  
  
  
  
• PrePrint: A Parallel Architecture for Interactive Rendering of Scattering and Refraction Effects
  Abstract—We present a new algorithm for the interactive rendering of complex lighting effects inside heterogeneous materials. Our approach combines accurate tracing of light rays in heterogeneous refractive media to compute high frequency phenomena, with a lattice-Boltzmann method to account for low-frequency multiple scattering effects. The presented technique is designed for parallel execution of these two algorithms on modern graphics hardware.
  
  
  
  
• PrePrint: Visualizing the Material Surface of a Living Human Brain
  We present a highly realistic visualization of the surface material of living brain tissue. Our goal is to improve the appearance of internal anatomy models used for education and training purposes, and our initial experiments have focussed on the human brain. Traditional teaching aids include images in text books, cadaveric dissection (where the appearance of tissues is quite different from when the tissue was still alive), plastic models (that are far from visually realistic) and computer generated models from medical images (which are mostly grey scale or use pseudo colours). In this case study we have been granted access to both an operating theatre during a neurosurgical procedure, and a dissection room at a medical school. The parameters required for a bidirectional reflectance distribution function (BRDF) are obtained and then used in the rendering process. Visualization of the brain’s surface is achieved in realtime by utilizing the GPU, and includes support for ambient occlusion, advanced texturing, sub surface scattering and specularity.
  
  
  
  
• PrePrint: Freestyle Group Formation Generation for Agent-based Crowd Simulation
  Relatively few algorithms take collective group features such as group formations into consideration when simulating a crowd. Recent approaches often manually specify hard constraints such as pre-defined key-frame agent formations and interpolate them during the simulation. We propose an interactive and scalable framework to generate freestyle group formations and transitions via natural and flexible sketching user interactions. Our group formation generation algorithm can automatically compute a plausible agent distribution in the target formation and agent correspondences between key-frames. On the other hand, our introduced two-level formation trajectory control scheme is able to guide the agents' moving/transition paths from an initial formation to the target formation with intuitive user control. Through numerous crowd simulation experiments, we found that our proposed framework is capable of efficiently generating various freestyle group formations in a crowd on-the-fly, while providing sufficiently flexible control to users.
  
  
  
  
• PrePrint: An Efficient Modeling System for 3D Scenes from a Single Image
  This paper presents a system for assisting the user to create a 3D model easily and quickly from a single image. Our scene model is composed of a background and foreground objects whose coordinates are calculated based on a “boundary” between “ground” and “wall”. Moreover, we introduce a fast method for extracting a foreground object by combining image segmentation and graph cut-based optimization. We show that the proposed system enables efficient modeling of foreground objects, easy creation of their textures, and rapid construction of a 3D scene model that is simple but produces sufficient 3D effects.