In a television system in which detail information is multiplexed with a carrier and transmitted over on NTSC channel, the picture-quality on existing receivers is degraded by the disturbance caused by the multiplexed component. we propose a method of "shuffling" by which the phase of the multiplexing carrier is shifted on a line and field basis in order to reduce the disturbance. The locations of the carriers with various shift patterns in three-dimensional frequency space and the characteristics of the human visual system are discussed. subjective assessment tests are carried out in order to select the best shift Pattern. As a result, the method using the most suitably shuffled carrier reduces the disturbance to less than 50% for ordinary Pictures and 38% for some particular pictures.

View full abstract

Ordinarily, several oil-based paints are stirred to generate marble patterns. In this report, the process of Stirring paints are simulated to generate marble patterns. AS it is difficult to obtain fluid flow exactly, it is expressed in terms of velocity fields derived from stream functions. Changes of positions of particles of which fluid consists are calculated and time changing patterns are generated.

View full abstract

In nature, many things are complex in shape and movement. Realistic representation of human beings and animals is made using computer graphics, but there are few researches on the representation of human hair. We have already proposed a stochastic model of movements of threads to try to represent thread-like objects, including human hair, blown by the wind. For such cases, an appropriate way of choosing parameters like the wind direction and strength was not known. We now propose a method using the solution of the two-dimensional potential flow equation in order to give values to these parameters.

View full abstract

We have developed a parallel rendering system to generate high-quality HiVision (HDTV) images. This system uses a hybrid MIMD architecture which combined Shared-Memory approach with 2-D mesh connected multiprocessor design, consisted of 85 PEs based on the Inmos T800 32bit microprocessor. Rendering algorithms such as Z-buffer, Phong-shading. A-buffer, texture mapping, bump mapping, reflection mapping are implemented in firmware. This paper discribes the architecture of the system and parallel algorithms for rendering.

View full abstract

A high-speed image generator for high quality HiVision computer graphics has been developed After sending data from a host computer about virtual 3D world such as objects, lights, and camera. the image generator quickly generates a picture. NTSC output is available as well as HiVision signal. A mat signal is also available for picture composition in post predation stage. Two VTR control ports for frame by frame recording, and frame grabber for external HiVision and NTSC pictures are equipped This image generator makes it feasible to produce high quality HiVision computer graphics animations for various applications.

View full abstract

To synthesize images more comprehensibly, image processing techniques can be applied to computer graphics. One example, drawing and emhancing profiles and internal edges, is discussed in this paper. When a shaded image is generated, the range image can be easily obtained as a by-product. Since profiles and edges are discontinuity of the range image, they can be produced by using differential operators. Using this technique, artificial factors such as the strength of edges are separated from geometric or physical factors such as shapes, lighting conditions, etc. Therefore, this method is effective for trial and error to find the best enhancement condition. They can be also applied to draw or to enhance edges of complicated surfaces, or reflected or refracted objects.

View full abstract

Facial. image processing is very important for achieving better human-machine interfaces and intelligent video-communication systems. This paper presents a hierarchical modification method to synthesize human heads and facial images. The proposed method is composed, of two processes: (1) a global process to roughly modify the 3-dimensional head shape model, and (2) a local process to tune up each portion of the head (e.g. eye, nose and mouth). Very natural head images-have been obtained using-the proposed method.

View full abstract

In this paper we present methods for rapid hidden line removal for structures which consist of lumber. we can speed up the procedure, by restricting objects which require precise calculations of intersection based on intersection check of bounding boxes. In our method, for rapid geometrical search, we use MD-tree as a data structure. And to further speed up the procedure, we present 2 methods based on coordinate transformations by using properties of lumber: one is based on a oblique coordinate system and the other is given by roating the image plane.

View full abstract

This paper proposes a progressive coding scheme, named contour line coding, that involves hierarchical architecture of image data. Most of bi-level image coding schemes like DCPM handle image as a set of '1' and '0' where '1' means black pixel and '0' means white pixel. We propose an other description for it. It presents the image by it's contour. Because contour lines if degital image are consist of vertical lines and horizontal lines, only cornerpoints are necessary information for this scheme. Like outline font coding, this scheme can be used to image transformation efficiently.

View full abstract

This paper presents a facial feature. information extraction and face reproduction method for the Virtual Space Teleconferencing System, and Shows the experiment results. In this System, there is a need to generate three-dimensional facial models from facial images and to synthesize images of the models virtually viewed from different angles. Feature points are extracted from front and Side views of the face with profile features and information about the position and correlation of the feature points. The system synthesizes images after modifying a face model with three-dimensional data from the extracted feature points. In this method, it is possible to generate a face model Without a human operator as in other conventional face modeling techniques.

View full abstract