In three-dimensional computer graphics, a set of polygons are generally used to represent a virtual space. In this situation, if the finer representation for the virtual space is required, the number of polygons constituting the virtual space is increasing, which gets rendering speed decreasing. In addition, to download through the Internet such a large number of polygons cost much communication time. To solve these problems, we use relief textures, textures with depth information, as distant view billboards in the walk-through system on a network.

View full abstract

In this paper, we introduce a new 3D video display scheme that can present motion parallax for dynamic events recorded in a large-scale space such as a soccer stadium, concert hall, etc. We propose a simplified 3D object representation method with a single 3D plane and 2D texture. This enables the reduction of data size and the video rate generation of observer's view with small latency. Even though the detail of each 3D shape is lost, the positional relation among multiple objects can represent the motion parallax. The static background is supplied as ordinary 3D CG model and it is combined with the moving foreground objects. Another feature of our scheme is that it can receive stream data of 3D video from a server via Ethernet. The encoding framework of 3D video data stream and the rendering technique of observer's view are also described. The feasibility of the proposed scheme was demonstrated by a prototype system implemented by PCs.

View full abstract

Radiosity methods make it possible to render photorealistic scenes in interreflective environment. However, calculation time for rendering a scene increases, and it makes difficult to generate computer animation, especially such the scene where a light source is moved, because each frame of animation is rendered from scratch. To solve the problem, we propose a new approach. Illuminance distribution of a light source in between two different light positions is calculated quickly by warping isophotic lines that are generated from each illuminance distribution of the two light source positions. The usefulness of the proposed method is demonstrated in rendering a scene under interreflective environment.

View full abstract

The CG technology for realistically representing the various things and phenomena that constitute natural scenery is becoming an increasingly important theme. The image generation of moss is also a very interesting theme from the viewpoint of visual simulation of both the natural texture and secular change of the ground and stonework, e.g. garden lanterns, stepping stones, paving stones and stone walls. In this paper, we propose a method for simulating the growth of moss qualitatively that takes into account several factors of the growth environment, such as amount of sunlight, temperature, humidity, and soil.

View full abstract

Volume visualization has served as an indispensable tool for peering into the inner structures of volumetric datasets. However, there exist a few reports on how to visualize effectively time-varying volume datasets, which arise frequently in a variety of numerical simulations and measurements. In this paper, the concept of volume data mining is extended to propose an environment, termed T-map, for visually exploring a large-scale time-varying volume dataset, where a pixel-oriented information visualization technique is employed so that the users are allowed to interactively specify partial spatiotemporal regions with characteristic change in topological structures prior to detailed and comprehensible volume visualization processes. A case study with a dataset from atomic collision research is performed to illustrate the feasibility of the present method.

View full abstract

This paper presents a robust mesh-watermarking algorithm that modifies low-frequency components of the mesh's shape for watermarking. The watermarking algorithm described in this paper tries to alleviates two weaknesses of our previous mesh-watermarking algorithm¹⁾, that are, (1) the weakness of the watermarks against vertex connectivity changes, due, for example, to mesh simplification, and (2) the high computational cost of the algorithm that prohibits processing of larger meshes. We added resiliency against vertex connectivity changes by recreating the connectivity of the original mesh on the watermarked and possibly attacked mesh by means of mesh resampling. We reduced the computational cost by watermarking disjunct subsets of the mesh. As with our previous watermarks, our new watermarks resist mesh smoothing, cropping, and additive random noise. Furthermore, our watermarks resist a wider class of attacks, including, for example, an attack that combines mesh simplification, similarity transformation, and cropping. In addition, our new algorithm is capable of watermarking larger meshes having tens of thousands of vertices. For example, a watermark embedded in a mesh having 66K polygons survived an attack that combines a mesh simplification down to a half of the original polygon counts followed by a cropping and a similarity transformation.

View full abstract

In this paper, we describe a system for arbitrary view generation from multiple view images using octree and silhouette. Our system consists of four calibrated stereo cameras. Each camera is connected to a PC that locally extracts the silhouettes from the color and disparity captured by the camera. The four silhouette images and camera images are then sent to host computer to perform 3D reconstruction. For making the reconstruction faster, the object 3D space is represented by octree structure. We develop a new algorithm for fast speed constructing octree. The algorithm can reduce time complexity to check if a node should project 8 cube vertices to image plane, using a stack that keeps parents' temporary cube type. After reconstruction of model, we render arbitrary view images with coloring all voxels included the parts of surface in the model. By using our algorithm, our system runs in semi real time computation, (about 5 frames per second) for generating arbitrary view image.

View full abstract

The recent high advancement of CCD device is making quantitative image measurements practical in many areas. However, various kinds of distortions are included in the obtained image such as distortions due to the aberration of lens and inaccurate setting of camera posture (position and direction), which cannot be ignored for accurate measurement. In this paper we describe a method of correction of such distortions and treatment of digital image in the accuracy of sub-pixel order. Utilizing a high resolution CCD camera with more than six million photocells, we propose a correction method for image distortion caused by the lens' distortion and camera' s angle of view can be obtained. We also experimentally verified that, the proposed method could achieve sub-pixel resolution with 4 times fidelity compared to the digital resolution of the camera.

View full abstract

In this paper, we present a method of the correspondence of character areas to arrows for road guide sign recognition to develop a driver support system in ITS. Generally, on a guide sign, there are arrows representing directions and characters representing destinations and distances. First, arrows are detected using their shape characteristics. Next, the sign is classified using the number of arrow objects and arrowheads. Then the character areas on the sign are assigned to their corresponding arrows by segmenting the sign using arrowhead positions. We tested 65 images zoomed in signs. 154 arrowheads on these signs were detected completely. In 62 images, character areas were assigned to arrows successfully. Hence, the usefulness of our method was shown.

View full abstract

We have developed a visual inspection-system with high speed and accuracy in which the quick matching to a reference image and the judgments whether a product is good or defective are made under changes of a set of good or defective sample images. The system provides a fast pattern-matching function that yields a high-precision estimation of the object position, a robust product-acceptability decision function, and easy operations. The developed system can reduce inspection mistakes or oversights by human inspectors.

View full abstract

This paper proposes a new restoration algorithm for blurred images of which distortion process is unknown. The algorithm consists of these two processes. The first, we present a target image distortion process by using a suitable two-dimensional FIR filter. This expression is effective on which the distortion process can be presented by a suitable point spread function (PSF). Then, we find the coefficients of the inverse filter by using the real-coded genetic algorithm (RcGA) effectively.
Moreover, we proposed two effective convergence methods for RcGA. These are a new evaluation function definition based on the two-dimensional image spectrum, and a useful impulse noise cancellation method combined by a rolling-ball transform and a succeeding Gauss-type function operation.
We verified that the proposed algorithm has a capability to restore many types of blurred images almost perfectly by computer simulations.

View full abstract

We configured a binocular vision model reflecting some eye characteristics by considering and imitating the structure of the eyeball. We subsequently simulated and visualized both near and far vignette phenomena arising from the depth of field effect as well as diplopia in binocular vision. Furthermore, we applied this model to four phenomena that are well known in the field of visual psychology as visual illusion phenomena based on visual direction in binocular vision. From such applications, we obtained the same results as those in reports of psychophysics experiments. Because the proposed binocular vision model is configured based only on the geometrical relations between the two eyes and the eyeball lens characteristics, the illusion phenomena that we were able to confirm through these simulations can be reproduced using only the geometrical structure of the eyes and the eyeball characteristics, in principle. In other words, these phenomena can be explained without involving higher-order brain functions. Furthermore, the proposed model was able to correctly reproduce those illusion phenomena, verifying the adequacy of this model structure. Lastly, as a method for investigating visual functions, we examined the significance and role of engineering simulation using such a computer model that is presented in this paper.

View full abstract

In this paper, we propose a new data hiding method for image half toning. The proposed method achieves data hiding during the half toning process. Comparing with the other data hiding method for binary images, the proposed method can hide larger amount of data. For data hiding, we propose the block error diffusion method to hide multilevel values in image blocks. The binary image that is obtained by the proposed method can contain two through eight level value in each image block.
In the experiments, we compare the proposed method and the conventional data hiding half toning by the relationship between the amount of the embedded data and the quality of halftone images.

View full abstract

We propose a new creation method of stroke linkage information using priority management rule and simultaneous occurrence rule concerning concatenated strokes to develop practicable square style on-line character pattern recognition algorithm. The method divides beforehand a set of concatenated strokes into the pre-determined plural level according to its appearance probability, then generates a concatenated reference pattern using stroke linkage information applying the priority management rule which links reference pattern' s strokes until pre-determined stroke number in descending order of levels. On the process, the method applies the simultaneous occurrence rule which links all strokes belonging to the most significant level then generates only one concatenated reference pattern. These rules not only obstruct unnecessary concatenated reference pattern but also make automatic creation of stroke linkage information possible. In the experiment for daily use Kanji character pattern written in the square style, the method achieves practicable correct recognition rate in spite of less stroke linkage information than another method, and its effectiveness is confirmed.

View full abstract

We developed a half-open dual-dome screen (HODDS) system and evaluated the images. The algorithm for calculating the parameters of the HODDS system is shown. The stereographic resolution for the depth is measured as 16.27 chains/dome diameter, whereas that for the horizontal is 45.12 chains/dome diameter. The information entropy in HODD-type is measured; the entropies for the HODDS and a flat screen are 6.989 and 8.593, respectively. Easily comprehensible expressions for amorphous materials are shown in the form of molecular shapes and dihedral angles.

View full abstract