We present a real-time rendering method of translucent materials such as marble for large 3D models. To render translucent objects, it is necessary to take into account the subsurface scattering of light. However, rendering translucent objects with the subsurface scattering takes long time. In our method, virtual points are set around the object uniformly. The intensities of the subsurface scattering of light are calculated at each virtual point. The intensities of points that represent the translucent object are interpolated by the intensities of the virtual points. Our method can render about 2,000,000 points with the subsurface scattering in real-time.

View full abstract

[in Japanese]

View full abstract

Point-sampled geometry has become one of the most promising representations in computer graphics because, unlike conventional polygonal meshes, it can manage large-scale 3D data without accounting for the point connectivity. The common rendering technique for point-sampled geometry is to introduce a set of oriented disks called surfels, however, it still suffers from unexpected holes on insufficiently sampled regions. This paper presents a novel up-sampling method for filling such holes on the surfaces consisting of surfels. For filling the holes, the present method extracts gaps around each surfel and inserts new surfels to fill the gaps. Contrary to previous approaches on global optimization, this procedure allows us to place new surfels deterministically by referring to only its local neighborhood. The present method adds a smaller number of surfels by taking into account their overlaps together with the user-specified radii of surfels. Several experiments have been conducted to demonstrate that the present method can be applied to level-of-detail control of point-sampled geometry.

View full abstract

This paper presents a new method for resolution independent rendering of vector objects using a stencil buffer and transparency multisampling. Our method has no expensive CPU processes, and therefore it can render dynamically deformable vector objects efficiently.

View full abstract

Real-time rendering of dynamic scenes taking into account global illumination is one of the challenging tasks in computer graphics. This paper proposes a new precomputed radiance transfer method for rendering dynamic scenes taking into account diffuse interreflection between objects. The interreflection of light between objects is calculated by considering the objects as light sources where the illumination distribution on the surface of objects, are expressed by Fourier series. This makes it possible to render dynamic scenes efficiently with diffuse interreflection whereas this is not possible using conventional precomputed radiance transfer methods. We also propose a method that runs on GPU and demonstrate that our method can render dynamic scenes with diffuse interreflection at interactive frame rates.

View full abstract

Rendering of glares with computer generated images is a popular way to represent brightness of the scene. We propose a fast method to detect bright spots in the scene and put glare images using vertex texture fetch on GPUs. Our method can generate high quality glare images in real-time and is suitable for interactive applications like computer games and virtual reality.

View full abstract

[in Japanese]

View full abstract

Perspective projection distortion is exaggerated according to the feature of the rendering objects in hand drawings. We call such a projection nonperspective projection. It is necessary to project each local part of rendering objects by different camera in the nonperspective projection, so that, we use camera parameter field in 3D space as a data structure realizing such a projection. Furthermore, we present a method for editing the camera parameter field through appearances of objects on screen. As an application of the nonperspective projection, we demonstrate an example for fitting a 3D scene to hand drawing.

View full abstract

[in Japanese]

View full abstract

This paper proposes a method for interpolating motion data with multilinear models. A third-order motion tensor is approximated by low-dimensional core tensor using high order tensor approximation. The resulting motion is accurately synthesized by interpolating the principal components using geostatistics. Our method reduces computational and storage cost of motion interpolation.

View full abstract

[in Japanese]

View full abstract

We propose a new method for creating realistic animations of cumulus clouds by combining a fluid dynamics simulation and texture synthesis. Small scale dynamics and fractal patterns of cumulus are synthesized in according to the cloud states, which are derived from the atmospheric fluid simulation. Small scale dynamics is controlled by the buoyancy energy, and adapted to large scale behavior of the velocity field. In addition the density distribution is used for adaptive pattern synthesis. Various shapes are combined by the cloud density and fractal noise textures according to the cloud density distribution.

View full abstract

[in Japanese]

View full abstract

To meet highly aesthetic requirements in industrial design and styling, we propose a new category of aesthetic curve segment. Many of aesthetic curves in artificial and the natural objects are shown, by Harada et al, to be curves whose logarithmic curvature histograms (LCH) are represented by straight lines. In this research, we identify the overall shapes of aesthetic curves and clarify several characteristics of aesthetic curves. For interactive control, we present a method for drawing an aesthetic curve segment by specifying 3 control points and the slope of LCH.

View full abstract

The curve is the most basic design element to determine shapes and silhouettes of industrial products and works for shape designers and it is inevitable for them to make it aesthetic and attractive to improve the total quality of the shape design. If we can find equations of the aesthetic curves, it is expected that the quality of the curve design improves drastically because we can use them as standards to generate, evaluate, and deform the curves. The authors have proposed the general equations of aesthetic curves as such a standard. However the aesthetic curves expressed by the general equations are limited to planar curves and it is necessary to convert the curves into B-spline forms to guarantee the compatibility of the curves on existing CAD systems. Hence in this paper, at first we show the necessary and sufficient condition for a given curve to have the self-affinity and then extend the aesthetic curves into 3-dimensional space. Furthermore we propose two methods to approximate them with B- pline curves by the least square to minimize the positional errors and the conjugate gradient to minimize the curvature errors as well as the positional ones.

View full abstract

[in Japanese]

View full abstract

Aesthetic surfaces usually consist of free-form surface regions and form-feature regions. When interactive deformation techniques are applied to such surfaces, they are required to satisfy the shapes of form-features exactly and free-form regions approximately. This paper proposes a new framework for managing hard and soft constraints. We solve the combination of hard and soft constraints using the Lagrange multiplier method and sparse linear matrix solvers.

View full abstract

In order to improve the efficiency of the modeling work for cyber restoration of historical city, a house structure description language has been proposed and developed. Each house structure is described by a series of simple words that indicates house construction part's ID. Width and height of each part is based on 90cm as a basic unit. Because we noted that the standard of the main parts of houses had become settled in width of 90cm, and height of 180cm. It is shown that various kinds of houses can be efficiently generated by our technologies to restore a historical city Edo.

View full abstract

[in Japanese]

View full abstract

In this paper we propose the hair motion reproduction method for cartoon hair. Our method is capturing a motion data from hair motions by nature phenomenon, such as winds, in an existing cel animation, and then creating the "Anime-like" hair motion model by using physical equations that is estimated by the captured motion data. That is, this method enables users to apply an "Anime-like" hair motion to another character as if it exists in the same scene as reference animation.

View full abstract

In this paper, a novel car designing tool, which helps car designers to design cars efficiently and stimulates their artistic sensibilities, is introduced. We synthesize cars by analyzing various existing car body shapes and extracting meaningful attributes that represent the features of these shapes. First, we create a database that consists of a variety of existing car shapes. We analyze all shapes using principal component analysis (PCA), then, gain attributes that indicate the typical features of those car shapes. By manipulating the weight of each attribute in the PCA space, we can create car shapes. Second, by analyzing the features of specific car groups, i.e. manufacturer, genre, era, etc, we discover attributes that discriminate between these groups. We can generate a new type of cars by adding these features onto any existing car i.e. a BMW-like Mercedes.

View full abstract

We propose a system of visualization and analysis utilities for large tree structure (VAULT). The VAULT has the following three functions: (1) "overview", (2) "progressive tree representation", and (3) "display of the detailed information by interactive selection".In the "overview" of a large tree structure, a spatially efficient node layout and link display without cluttering are necessary. Some algorithms for realizing this function are compared. In the "progressive tree representation", the edge collapse operation is performed. In the "display of the detailed information by interactive selection", the concept of focus+context is incorporated in the system in order to obtain more detailed information by the interactive selection. We demonstrate effectiveness of the VAULT by applying it to the growing process of a large branch-and-bound tree.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

We propose the method to reappearance anisotropic reflectance property of woven fabrics and to represent the unique texture based on the mesostructure of surface. In this paper, we estimate the micro surface geometry of woven fabrics based on reflection analysis. We consider the weave structure and we calculate masking on the convex surface of weave. We apply masking model to Cook-Torrance model for generating reflection. The micro surface geometry of woven fabrics is estimated from reflectance measurement. The reflectance generated from our model compare with the measured reflectance. The experimental result shows the effectiveness of our model.

View full abstract

Image mosaicing is a method for partitioning an input container image into small image tiles, and offers its non-photorealistic representation. However, conventional mosaicing methods suffer from the absence of clear edges because they use a bottom-up scheme that first extracts local features such as object boundaries to constitute the whole image partition. This paper presents a robust method for composing quadrilateral image mosaics in a top-down fashion by respecting the low-frequency intensity gradation. The present method also allows us to incorporate the local features into the existing global quadrilateral partitioning by taking account of its dual network.

View full abstract

Video game machine was cheap but people drawing row-res video game graphics used them brain. It takes them a long time to draw row-res video game graphics. Video game machine have been powerful. For the reason, video game graphics become 3D graphics, but row-res video game graphics is effective for row-res display. We researched the method of drawing pixel art. Then we need to output binary contour lines method for pixel art. We down scaled binary contour lines. We propose coloring method for pixel art.

View full abstract

In this paper, we propose a method for generating colored-paper mosaic images automatically from an input image. We use the hierarchical Poisson disk sampling algorithm to determine torn colored-paper tile locations. Our method makes it possible to control overlapped area among neighboring disks. We also propose a method for casting shadows by overlapped colored-paper tiles to simulate real pictures of colored-paper mosaic images.

View full abstract

In this paper, we propose a non-photorealistic rendering (NPR) method for generating paper mosaic images from input images. We modulate the brightness of input image on the basis of the number of overlapped pieces of papers and texture of papers. We show some paper mosaic images generated by our method.

View full abstract

There are not many studies on painting tool which is focusing on oil paint materials. In this paper we propose a new viscous model for oil paints which is based on Lattice Boltzmann Method (LBM), one of the fluid simulation methods. This model has advantages that strokes have realistic profile according to oil density and time after the paint to previous oil paint models. It will contribute to the generation of rich strokes on paint tools.

View full abstract

Isosurface is a set of points where scalar values are constant, and it generally forms surfaces. Simultaneous display of multiple semi-translucent isosurfaces on a screen is useful for visualizing large-scale 3D data. It realizes comprehensive representation of the 3D scalar fields, as well as volume rendering techniques. In this paper, we focus on visualization of 3D scalar field composed of volume data by means of semi-translucent multiple isosurface plots. A new method for displaying semi-translucent multiple isosurfaces will be proposed to fulfill realtime rendering of scalar fields. ISSUE is a known algorithm to quickly generate isosurfaces. ISSUE uses 2D span spaces where minimum values of elements are assigned to horizontal axis, and maximum values of elements are assigned to vertical axis. In our algorithm, elements of volumes are placed onto a 3D span space. The third axis is the distance from the view point. Our algorithms can quickly search for elements intersecting with isosurfaces using the span space in the descending order of distances from the view point, and realize representation of semi-translucent multiple isosurfaces.

View full abstract

In this paper, we have developed a method for directly rendering cylindrical volume data. The proposed method employs a splatting method where we design an ellipsoidal kernel and a footprint with an elliptical shape. To efficiently render cylindrical volume data, only a small number of footprints is generated, making use of both characteristics of a parallel projection and symmetry of the kernel alignment in cylindrical volume data. We determine the projection order of ellipsoidal kernels using regularity of voxels in cylindrical volume data. Visualizing a fundus oculi reconstructed from cross-sectional images scanned radially with an “OCT ” demonstrates the usefulness of the proposed method.

View full abstract

Image Analogies algorithm performs a search of the tree to find a similar pixel in the training images. GPU’s architecture is unfit for the trees which require using stack or queue when a search is performed. To solve this issue, our method builds a quad tree by using K-means. The tree doesn’t need back-track to be searched. Hereby, the tree is suitable for GPU and we have achieved to accelerate Image Analogies with GPU.

View full abstract

In computer graphics, the simulation-based method is widely used to represent the realistic motion of fluid. In fluid simulation with free surface, it is important to track the free surface evolution precisely. Although the level set method is common as the surface tracking method, naive implementations suffer from numerical dissipation. To address this problem, the particle level set method is proposed, but its computational cost is high to improve the precision. This is of a problem especially for the large scale simulation. We focus on the graphics hardware which continues to improve the performance, and propose the method which accelerates the particle level set method by using its parallel computational power. We demonstrate that the proposed method tracks free surface faster than the previous methods with the same precision.

View full abstract

This paper presents evaluation results for two GPU implementations that aim at accelerating mathematical morphology. Our methods decompose the computation into independent pieces in order to achieve parallelization on the GPU. We have two different implementations, namely the scatter-based and gather-based methods, each employing one of two decomposition schemes. The scatter-based method parallelizes pieces, each associated with a pixel in the structuring element. On the other hand, the gather-based method decomposes the computation according to pixels in the original image. The latter method is an extended version of Eidheim's method, which is capable of dealing with arbitrary structuring elements. The experimental results show that the scatter-based method is seven times faster than a naive CPU implementation for a structuring element of 100 x 100 pixels. The gather-based method achieves the highest performance but fails to deal with structuring elements larger than 40 x 40 pixels, due to the limitation on the number of instruction slots.

View full abstract

In this paper, we propose methods for real-time visualization of 2D numerical simulation on GPU. A scalar field is visualized with a distribution map and a contour line map, and a vector field is visualized with arrows and particles. To real-time visualization of arrows, we propose the new technique having used Vertex Texture Fetch (VTF). A vector value is fetched by VTF from simulation results to the polygon prepared beforehand. The coordinates of a polygon are transformed based on it, and the arrow texture is mapped on a polygon. Thus, an arrow can be drawn on real-time.

View full abstract

Recently researches on techniques to use simulations based on the physical laws for CG animations are becoming popular and popular. Although paper exists ubiquitously around us in daily life, it is not straightforward to express its folding lines, fractures, and textures and no much effort has been made on researches on paper for CG. Hence the purpose of this paper is to simulate paper fracture based on the physical laws and use the results for CG animations as the first step to the CG-oriented paper simulation. The fracture simulation of paper is performed at first by modeling a piece of paper with triangular meshes and applying the FEM to calculate their stresses and strains. Then we determine the starting position of the fracture and the direction of the crack and rebuild the structure of the meshes to express the fracture cross sections. Furthermore for the cross section originally expressed by piece-wise linear segments, by use of the turbulence function we generate natural-looking geometrica ly complicated cross sections as textures and map them to the meshes to represent the paper characteristics that become clear when the paper is broken. The geometry of the fracture cross section remains similar even though the mesh size is changed because of the noise-mixed texture mapping, that allows us to use relatively large-sized meshes and simulate the paper fracture in almost real time.

View full abstract

Recently, simulation of natural phenomena is one of the important research topics, and many methods to render water, smoke, cloud and flame have been developed. In this paper, we focus on the simulation of flames among these natural phenomena. First, we propose an interactive method to create flames. Our method allows the user to control the shape and motion of the flames. This is achieved by combining cell automaton method and particle system. Secondly, we propose a fast rendering method using wavelets for the flames and surrounding objects illuminated by the flames. This method realizes the real-time simulation of dynamic flames and the real-time rendering of the objects illuminated by the flames.

View full abstract

In this paper, we propose a real-time visualization method which keeps track of critical points to visualize the dynamics of a 2D physical field adaptively. We show the effectiveness of the method with an application to visualizing datasets obtained by the measurement-integrated simulation in the hybrid wind tunnel.

View full abstract