The action of grasping an object is very important for an autonomous character agent. When an agent grasps an object, it has to determine which portion of the object to grasp. However, the selection is not easy, because there are a lot of possible grasping methods for one object. Our method consists of the following steps. First, portions that can be grasped on the surface of an object are detected as candidates for a grasping position. Then, hand shapes corresponding to each candidate's shape are presumed using a neural network. Therefore, their candidates for a grasping position are evaluated by stability using robotics technique.

View full abstract

[in Japanese]

View full abstract

In this paper, we propose a novel representation and the estimation technique of ROM (range of motion) from Mocap (motion capture) data set, suitable for human animation. Although, there have been several works for ROM in character animation field, none of them can deal with the interrelation between joints mainly caused by biarticular muscles. ROM measurements in medical contexts are limited for rotations on three reference planes. And more, these measurements does not always give as proper value for human animation, because of the disagreement between human skeleton and the bone model on the computer. In our method, we defined the probability density functions, gaussian kernels, as ROM, derived from Mocap data which has the same bone model as the final output. Moreover, taking advantage of kernels, our method enable us to approximate ROMs considering the interrelation between joints.

View full abstract

Realistic rendering methods of natural phenomena in real-time have a variety of applications, such as flight simulators or computer games. The scattering effect due to atmospheric particles is one of the most important elements in creating realistic outdoor images. Shafts of light are caused by objects which shut out the sunlight, such as clouds and mountains. This paper proposes a method for creating outdoor images including the sky color and shafts of light. In our method, shafts of light are displayed by subtracting the intensity of the shielded light. Our method is accelerated by using a fragment shader, which is a function of the latest graphics hardware.

View full abstract

[in Japanese]

View full abstract

Recently, realistic image synthesis using computer graphics has been used in various fields. The reality of the synthetic images can be improved by taking into account specular reflections such as mirrors, sufaces of cars, glass windows of buildings reflecting other objects. So, there are many methods to simulate the specular reflections. However, the simulation of this phenomenon is quite complex and therefore is very time-consuming. This is especially true for glossy specular reflections. On the other hand, the recent advancement of graphics hardware has made significant progress in the performance. This encourages researchers to develop methods for accelerating image generation processes using graphics hardware. We present a fast method for rendering glossy specular reflections considering reflectance distribution functions. In the real world, different parts of an object have different reflectance distributions. Our method can render this kind of objects.

View full abstract

Viewpoint selection is an important task because it considerably influences the amount of information contained in the 2D projected images of 3D objects. Although several methods have been proposed that calculates the optimal position of the viewpoint especially for surface meshes, none has been done for solid objects such as volumes. This report therefore presents a new method of calculating such optimal viewpoints when visualizing volumes using ray-casting techniques by extending an existing method for surface meshes. The major idea behind the present method is to decompose an entire volume into some set of components and then find a globally optimal viewpoint by computing a locally optimal viewpoint for each component. Furthermore, the method employs transfer functions to assign a weight value to each component so that it can emphasize individual features embedded in the volume using the calculated viewpoint.

View full abstract

This paper proposes a simple and intuitive user interface for volume segmentation. The user traces the contour of the target region using a 2D free form stroke on the screen, and the system instantly returns a plausible 3D region by applying a segmentation algorithm. Our system first computes the 3D location of the user-specified 2D stroke based on the assumption that the user traced the silhouette of the ROI, that is, the curve where the gradient is perpendicular to the viewing direction. The system then places constraint points around the 3D stroke to guide the following segmentation.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

Our purpose is to transform an input line into an oriental brush stroke automatically. However, there are many parameters required for the rendering, and it is not simple to determine them from input lines. In order to solve this problem, we use a continuous HMM which is one of the machine learning methods. We prepare the captured stroke data as training data, and then train a continuous HMM. We can obtain an oriental brush-like painting which is based on physical characteristics of brush and a drawing style by applying the trained continuous HMM to an input line.

View full abstract

This paper presents a computational method for occlusion-free animation of geographical landmarks and its application to a new car navigation system in which driving routes of interest are always visible. This is achieved by animating surperspective projection images where geographical landmarks such as mountain tops and roads are projected as seen from different viewpoints. The proposed approach consists of: extracting geographical landmarks automatically from the 3D terrain surface; deforming the 3D terrain surface using 2D screen constraints imposed on the landmarks; and retaining temporal coherence in the animation by controlling the deformation of the terrain surface. Several occlusion-free route animations using real data of a mountain area are created to demonstrate the effectiveness of the proposed method.

View full abstract

In this paper, we discuss NPR for a monitoring camera. Monitoring cameras take overshoot or undershoot images because of various lighting situations. In order to show human facial features clearly, it is necessary to compensate these images. However, general image compensation techniques require manual parameter settings for each lighting situation. Therefore, we apply NPR to these obscure images in order to emphasize the features. Our experiment shows that the video images by proposed method are superior to original video images and compensated video images by conventional methods in recognition speed.

View full abstract

[in Japanese]

View full abstract

In this paper, we propose a visualization method of melting ice by calculating a thermal field by airflow. We create two calculation fields; an Air Calculation Field (ACF) has a property of fluid, and an Object Calculation Field (OCF) decides the phase transition by thermal energy diffusion. We achieve the thermal energy propagation between the ACF that surrounding the OCF and the OCF. The thermal field updating with Time-Step arises herewith, and we create melting animations.

View full abstract

[in Japanese]

View full abstract

In recent years, attention has been paid to particle based fluid simulation, with several methods being developed to incorporate particle based simulation into CG animations. These methods reconstruct water surfaces that are usually represented by polygons. However, the computational cost of the surface reconstruction is quite high. Therefore, it is difficult to render the result of the particle based simulation at interactive frame rates. To address this, we present a real-time rendering method for particle-based simulation, taking into account optical effects such as reflection, refraction, and caustics. We present an efficient GPU accelerated surface reconstruction method from particles.

View full abstract

This paper proposes an efficient technique for rendering a scene with participating media. To render a scene with participating media, radiance at many sample points should be calculated. In our method, we first approximate the in-scattered radiance. Then, by using the precomputed radiance values, importance sampling method is used to render a high quality image.

View full abstract

[in Japanese]

View full abstract

The analysis and synthesis of human skin color is important in many different application areas. The skin color depends on some histological vatriables such as the concentration of pigments of melanin and hemoglobin in skin layers. The present paper describes a method for estimating surface-spectral reflectance of human skin based on an optics model and applying the estimates to 3D realistic image rendering for a part of the human body. The human skin is modeled as the two layers of turbid materials for epidermis and dermis. An estimation algorithm is then developed using the Kubelka-Munk theory to the two-layer model. The parameters representing the concentration of pigments are determind based on spectral-reflectance measurements of human skin surface. In the application step, we describe a technique for rendering realistic skin images of a skin surface as a 3D object. In experiments, the accuracy of the estimated reflectances is shown, and skin color images are created under a variety of conditions on illumination, viewing, and pigmentation.

View full abstract

Tomogram of eyeground images by Optical Coherence Tomography(OCT) are a set of the echo information on a minute portion. These images contain many random noises and gap between images.Images of OCT are strongly influenced of refraction, pulsation by the living body organization. Therefore, it is difficult to form 3D volume images. In this research, a noise removal of 2D images and position alignment processing were performed as a pretreatment of 3D visualization. 3D images contribute to medical treatment of eyeground disease and informed consent.We utilized a texture-based volume rendering technique, and evaluated the usefulness of the 3D eyeground model in the medical treatment of eyeground disease.

View full abstract

In this paper, we propose a fast and efficient method for real-time ray-casting with less memory for irregular grids, which we call tetrahedral adaptive grids. Since the quantity of volume data reconstructed from CT or MR images are generally very large, hierarchical volume models capable of compressing data are required to handle several tasks efficiently. The tetrahedral adaptive grid have been proposed for generating a hierarchical tetrahedral model by recursively subdividing a tetrahedron according to local field properties such as gradients and curvatures of isosurfaces of input volume data. Experimental results on an artificial data demonstrate the performance of our proposed method.

View full abstract

[in Japanese]

View full abstract

Color is very important in setting the mood of images and video sequences. For this reason, color transformation is one of the most important features in photo-editing or video post-production tools because even slight modifications of colors in an image can strongly increase its visual appeal. However, conventional color editing tools require user's manual operation for detailed color manipulation. Such manual operation becomes burden especially when editing video frame sequences. To avoid this problem, we previously suggested a method that performs an example-based color stylization of images using perceptual color categories. In this paper, we extend this method to make the algorithm more robust and to stylize the colors of video frame sequences.

View full abstract

We propose an acceleration method for fragment-based image completion [Drori et al. 2003]. The fragment-based image completion can synthesize a visually plausible image. However the time complexity of this method is O(n^2). We exploit Gaussian pyramids to improve and accelerate the algorithm. Moreover, to prevent the deterioration of resulting images caused by Gaussian pyramids, texture classification maps are introduced for specifying similar texture regions. The combination of Gaussian pyramids and texture classification maps allows fast image completion while preserving visual plausibility. This method can synthesize a fine image, e.g., one Mega pixel, within a minute.

View full abstract

Recently, many researchers have paid attention to a technique called non-photorealistic rendering (NPR). In this paper, we propose a method for painterly rendering of water surface by making use of techniques developed for three-dimensional computer graphics. We focus on the non-photorealistic rendering of vortices that are one of the important phenomena observed in the water surface. Our method makes it possible to create animations of water surface including generation and extinction of the vortices. For NPR, it is very important to reflect the user's intention on the resulting animation. So, we develop a system using the proposed method so that the user can control the motion of the flow and the vortices. By using the proposed system, the user can create NPR animations of the water surface taking into account the water flow and the vortices.

View full abstract

In this paper, we present simple cellular model for computer generated dyeing. A cloth is represented as two layerd celluar model.The weft and warp are set as positional relationship of strings of cloth. Each cell has transportability and maxmum capacity of dye. We improve a simulation of the cappillary flow model of paper and watercolor. Because the cloth has weft and warp, we give 3 cases of dye transport. This model bears a close resemblance to real stain in both coloristic change and mottled appearance on the fringe.

View full abstract

[in Japanese]

View full abstract

When novices at drawing are going to make an illustration in mobile environments, there is a problem that they cannot handily get a beautiful curve which they want to draw. This paper proposes a support system for a novice to make an illustration easily. The system has the three functions: 1) A freehand drawn curve which has extra fluctuations is revised into a smooth curve; 2) Partial modification can be made by the following method: one or more modification candidate curves are input, and a part of the former curve is revised with one of the modification candidate curves that is chosen by the user; and 3) Two curves are connected smoothly with the rest of the former curve keeping the form as much as possible. The experimental result showed that the proposed system was effective to novices at drawing.

View full abstract

We have proposed a computer-graphical method to realize the head of a robot with several kinds of sensor to respond to a peripheral environment. There, it has shown that the face and hair of a robot can be displayed realistically according to a peripheral environment such as light, wind, and the pressure of a finger. This paper proposes a method to represent the eyes of a robot in the same way. Combining CG with a camera, a prototype system shows that the eyes of a robot changes according to a surrounding environment realistically.

View full abstract

We purpose a method to create human walking animation in the arbitrary direction with little cost. We apply image-based rendering techniques for the cost reduction of a human modeling. Human walking sequence along a circle is captured as input video images. A human image in each walking phase and each walking direction is extracted from the video. By interpolation of these images, the human image walking in the arbitrary direction is created. In this research, the optimal interpolation is obtained by matching each scanline blocks.

View full abstract

The subject described herein is a novel method for modeling and visualizing the deformation of a rheological object. We propose a dynamic adaptive triangle mesh of which the vertex link is a rheology element consisting of the group of a viscous element joining to two viscous and elastic elements associated parallelly. Based on the physical and geometrical properties of a rheological object, the mesh conforms dynamically to the deformation. We apply the dynamic adaptive triangle mesh for the problems of (i) visualizing force interactions of a deformation process, and (ii) rendering of cutting of a rheological object. In the simulation of a cutting phenomenon, the size of a mesh element newly arising in a cutting path is always possibly smallest; and hence, the cutting paths can be visualized finely in detail, despite the fact that the element number (the number of mesh vertexes) of the model is small in comparing to that of the modeling methods using non-adaptive meshes.

View full abstract

In order to create realistic animations of computer graphics for movies and other commercial films,the physically-based simulation is mainly adopted to realize gas or liquid flows for rendering flames and water movements.However no babble is dealt even for simulation of a low viscosity liquid like water and it looks like a high viscosity liquid rather than the real water.Hence the purpose of this research is to realize a simulation of fluid with babbles that can treat their deformations as well as movements.In this research, we use non-uniformly subdivided grids with an adaptive data structure instead of uniformly subdivided grids to decrease the amount of data and the computation time required to guarantee the same accuracy, and we make the size of the grids around babbles smaller than other parts.We compute pressure and velocity field with using the adaptive grids generated in such a way and develop an algorithm to update babbles' velocities by use of them.Compared with the simulations with the uniformly subdivided grids, those with the adaptively subdivided grids require about one tenth memory size and one sixteenth computation time.

View full abstract

[in Japanese]

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. Harada et al. insist that natural aesthetic curves like birds' eggs and butterflies' wings as well as artificial ones like Japanese swords and key lines of automobiles have such a property that their curvature logarithmic curvature histograms (LCHs) can be approximated by straight lines and there is a strong correlation between the slopes of the lines and the impressions of the curves. In this paper, we define the LCH analytically with the aim of approximating it by a straight line and propose new expressions to represent an aesthetic curve whose LCH is given exactly by a straight line. Furthermore we derive general formulas of aesthetic curves that describe the relationship between their radiuses of curvature and lengthes.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract