600Mbit/s circuit switching system which provides mainly the moving-picture service is proposed. This system exchnges digitized NTSC and HDTV signals and integrates the end-to-end communication, center-to-end communication and video distribution services. The communication service or distribution service is selected by the layer three signaling of the I interface. Video channels are switched by control of the higher layer of the subscriber signaling. This system is suitable for a building block architecture. The core module accommodatesfrom tens up to about three hundred subscribers. All of the modules have load and function shared control and standalone functions such as subscriber access, trunk access,inter-office signalingand system management.
Effectiveness of visual communication can best be achieved if the observer be-comes totally involved in the displayed image through a feeling of "being there". This sensation of virtual existence or enhanced reality is activated by the capability of a precise pixel expression and a large visual field display. However, these two requirements are contradictory because view angle per pixel deteriorates as the size of the visual field increases. To solve this problem, a new large visual field display system, based on the fact that human vision is most acute over a narrow area centered on the point viewed, is proposed and its construction is shown. The system consists of a large visual field image and viewed region image. With an eye movement tracking instrument, the gazing point within the large visual field image is measured. And, a viewed region image from an area in the vicinity of the gazing point is optically superimposed on a large visual field image. Next, subjective experiments concerning the limitation of spatial frequency in the peripheral vision region are performed. The result shows the effectiveness of the proposed system.
A color information and reproduction method for the Virtual Space Teleconference System is described. In this system, different perspectives of a human image corresponding to different viewpoints have to be synthesize from a single input human image. The shape, motion and color information of the input image is first analyzed, and then a synthesized image is generated to suit the required conditions. The proposed method is concerned with the color information analysis and synthesis part in that process. The input human image is segmented into perceptual color regions of similar class, where the color information is coded independently of shape. Then the colors of the images corresponding to different viewpoints can be reproduced as required.
This paper describes a design of a full-color a component of a personal multimedia codec which is teleconference terminal. This codec transmits both color still pictures and narrowband ISDN (Integrated Services video over a Digital Network). These images are integrated with high handwritten images and resolution bilevel graphics. handwritten pointers for displaying on a single monitor.
Transmission of Hi-vision imagery which contains enormous amount of video information requires extremely high speed communication channel. This fact has been one of main barriers of possible Hi-vision application in various fields. In this development of codec, displaying details is considered a key feature of this Signal format and frame rate is traded for keeping resolution at a limited channel rate. It is expected to be an useful approach when transmission of high motion is not necessary.
Standardization for highly efficient video coding system is in progress at CCITT. In this paper, loop and post filters used in the video coding system are studied to be contributed to the standardization work by the Specialist Group on Coding of Visual Telephony, which is substantially responsible for making the recommendation. Through previous study for n×384kbps codec standardization, a simple loop filter with 3×3 pixel window has been proven to show good performance with a reasonable hardware cost at 384kbps. In lower bit-rate operations, however, some improvements on the loop filter have been felt required because noticeable block-shape and granular noises have been observed in the reproduced pictures. Aiming at improving the reproduced image quality, extensive experimants have been carried out. From the experimantal results, the followings have been obtained: (1) Considering perfomance, hardware cost and IDCT mismatch condition, the 3×3 low-pass filter is the most suitable as loop filter in 64-1920kbps video codec specification. (2) In lower bit-rate operations, an adaptive noise filter for post processing significantly improves reproducesed image quality.
This paper proposes a new design of a TV camera optical system with LED illumination in order to extract efficiently a pupil image which is an effective feature point for non-contact eye movement detection. The brightness distribution of the pupil image is calculated by raytracing using an approximate model that is expressed by the combination of 2 transparent spheres. It was proved that such simulation is effective by comparing the experiment results with the calculated results. Two conditions of picking up bright pupil images and dark pupil images became clear. By obtaining the differences between 2 kinds of pupil images, the pupil cart be accentuated against the background. Because the S/N ratio of the differing image is extremely high. The pupil is extracted by a simple thresholding process.
This paper presents a new region splitting method which uses perceptual color system. The algorithm is constructed from two major steps. First step is the region segmentation by the feature of the input image. The image is devided into three parts, one has the regions with high satulation, another is with low, and the other is with no satulation. Second step is the area extraction that is suitable for their features. At the first part of the image, the areas are extracted by using perceptual color system, especially hue information. Physical color system, like RGB is used for second part. In the last part, we use the lightness data.