We analyzed the variation in unconstrained handwritten characters using "variation entropy". As a result, it was difficult to find a common variation tendency for all characters. However, we found that for each character type the variation in unconstrained handwritten characters correlated with that of constrained characters.

View full abstract

Icon is constructed with some objects which have the nature of affordance. But we do not know exactly what objects we should use to make an icon set that represents and expresses certain functions. So we examined some objects in association with affordance of icon functions.

View full abstract

We tried to apply the back propagetion network(BPN) to recognize men-and-women and each Individual person by patial areas of front-view face (partial Faces). The experiment is performed by using two faces, including and excluding hair. Hereby, we investigated in which part of face is the most important feature.

View full abstract

The Multi-channel mechanism in color vision was experimentally studied. Channel characteristics for the color combination of red-green were measured using a method called a masking method. The existence of four channels for normal subjects and three for a deuteranomalia subject was found.

View full abstract

An eye-head control system is discussed. The share of the head becomes constant at a greater target distance along every meridion. The neck always tilts with the movement of the gaze. This result suggests that the visual system calculates both the horizontal and vertical displacements of eye and head while considering neck tilt.

View full abstract

Our purpose is to clarify the roll of viewing time while fixating on a certain point of an image. A new sight-line displacement analyzer which limits the subject's gaze time by inserting a masking pattern in his focusing area has been developed. The experiment task consisted of counting a patten embeded in the distractors. Starting time of masking after fixation [TM] was varied in five ways (66,134,200,270ms, without masking). The mean fixation duration during counting which corresponds to the processing time of counting is about 350ms, irrespective of TM, Even in 66ms of TM, counting accuracy exceeded 65%. This suggests that the retinal image can be acquired in less than 66ms.

View full abstract

This paper discusses the study of human eyes'fusion range of binocular parallax picture emplaying red and green random dot stereograms. The experiments we conducted show that the fusion range of human eyes'binocular parallax is directly proportional to the figure angle of the binocular parallax region in the stereo picture. For the people having rich the experience of stereo vision, the fusion range of the human eye's negative parallax is bigger than that of positive parallax. The fusion range of negative parallax can be increased through the accumulation of stereo vision experience, in other words, there is a learning process. In the paper, we also suggest, for complementary coloured (red and green) 3DTV, the fusion range of binocular parallax : |Δθ|<0.02arc

View full abstract

A relation between the binocular fusion and the target size is examined using wide field random dot stereogram. As a result, a disparity of binocular fusion increases with the enlargement of target size. However, the size of background does not influence the binocular fusion.

View full abstract

Miniature eye movement is characterized by it's randomness, small amplitude and unconsciousness, therefore, for a long time, this has been regarded as noise. In this report, we analyzed fractal dimensions of the difference between both eyes horizontally and vertically, respectively. We obtained around 7 as fractal dimensions of noncooperative components for each direction. This shows the possibilities of a higher dimensional chaos in our visual perception.

View full abstract

A concept of Line Spread Function (LSF) in depth was introduced. It was measured as depth discrimination function using moving bar-target depth perception system. The OTF in depth was derived from the results and compared with previously reported data which was obtained by field sequential stereoscopic TV system.

View full abstract

To clarify the interaction between right- and left eye controlling mechanism, we measured subjects' binocular eye movement and stereopsis, presenting a moving stimulus to one eye only. We found there was a case to percept motion in the depth direction when a convergence eye movement occurred. However, this relationship is not simple, and more complicated mechanisms may exist or these mechanisms may exist independently.

View full abstract

We studied on the objective measurements of the "Sensation of Reality" by utilization of the postural control response. We measured the subjects' body sway induced by the rotating 3D and 2D-image. Results have showed that the 3D-image has a greater effect on the subjects' postural control system than the 2D-image.

View full abstract

The Video Browser is a method of scanning a video program and generating a hard-copy index by which the editor can immediately see where scene changes occur and can quickly get a feeling for the overall organization and rhythm of program.

View full abstract

We have developed a new simulation system to evaluate today's various video systems (shown as Figure 1). It performs real-time signal processing with a parallel arrengement of 1024 DSPs(Digital Signal Processors). This simulation system features user friendly software environement, so we can easily design and examine the objective systems.

View full abstract

We have constructed a friendly graphical user interface(GUI) for a video signal processing system. Necessary the system operations such as parameter setting process ordering can be also controlled through the GUI. The system with GUI successfully simplified the laborious simulation planning process.

View full abstract

We designed a model of AV (Audio & Video) system. The model is separated by SYSTEM DATA into a construction phase and an operation phase. A system constructed according to such a separated model has many merits. Those are flexibility, high faculty, expansibility. We could have proven that by constructing two AV systems at exhibition rooms of SANYO in Tokyo and Osaka.

View full abstract

A Hi-Vision Gallery System, an example of AV system consisting of Hi-Vision equipments (Hi-Vision Videograph System, Hi-Vision Disc Player, etc.), has been recently developed. Easy operating, accepted explanation of how to operate with voices and sound effects by machines controled with MIDI, was appreciated.

View full abstract

[in Japanese]

View full abstract

In this work we propose an Optical Model as an altemative to display 3D-volume data, while the original values remain unchanged. The general principle of the new method proposed here, consists of two parts. First the volume data is mapped to some optical constant. Then using a plane wave for the illumination, and considering the boundary conditions stated in the Maxwell's equations together with the absorption and diffusion concepts, the volume data is visualized.

View full abstract

Holographic pattern measuring system is a combined system of a holography and a graphic image processing technique. A distribution of microscopical displacement can be measured, and it can be analyzed as the three-dimensional graphic image. However, in the system, the interferometry data have been treated handly, so both the error and the time loss were inevitable. In this study, the digital image processing technique was applied to detect automatical from interferometry image. As the results, more acculate pattern data coud be clearly and speedily detected

View full abstract

We identify two independent forms of error in stereoscopic vision : imprecision, which results from digitization and imperfect knowledge of the system geometry and location, and uncertainty generated by noise and the "correspondence problem". Although it is intuitive that both should decrease as we accumulate the evidence provided by multiple stereo-views, the lack of a solid mathematical framework for representing imprecise and uncertain information has hampered progress in this direction. We propose such a framework here, known as the Random Closed Set theory, and we show how it can be used for stereo-vision. Results of preliminary experiments on real images are presented.

View full abstract

A modified segmented Hough transform method which combines the local features of each pixel is presented for hierarchical feature extraction from a binary character pattern. Using this method, the precise stroke features of handwritten characters can be obtained.

View full abstract

In this paper, a feature extraction in facial image using 3-D wireframe model is proposed. Description of the extracted feature by parameters in three-layer-space is also under study.

View full abstract

This paper presents a facial image processing method in extraction and recognition of face component regions. After hair and face regions are segmented from human images, the face component regions are obtained using information of position and color of the components independent of face tilt and rotation. The processing method are useful for feature extraction from human images.

View full abstract

Abstract This study is an early works to 3-dimensional motion compensated predictive coding. We propose an available method to compensated the motion of a cube in behalf of polyhedron. This method is characterized by the object outline extraction and motion information ditection.

View full abstract

This paper shows that texture analysis of static images can apply to grouping and segmentation tasks in motion analysis. Motion is described in static images as blur or line-attaching images on moving objects which call "motion texture". A motion texture image is created by accumulating frame after frames the time and spatial fluctuation of the gradient gray level at each local area. We create motion texture images from random dot animations. When the density of motion vectors extracted from all parts of 2 succesive frames was set constant, the global perceputual impression differed from animation to animation. The difference impressions of global motion perceived in the animations may be explained by analyzing the motion texture elements. Furthermore, the grouping of moving objects can be successfully achieved by texture segmentation.

View full abstract

In this paper an improvement of gradient method for motion estimation is described. In some case incorrect vectors are estimated in the areas of a little spatial gradient or edge of a moving object. The non-linear processing for the spatial gradient is effective for this probrem.

View full abstract

The reliability of motion vector is calculated using the distance of video signals between two successive fields defined in orthonormally transfered space. A high correlation is found between this distance and the error in detected motion vector.

View full abstract

Possibility of picture quality improvement by transmission of chromaticity signals when chromatic signals are band-limited in narrow range on transmission routes, is studied. A chromaticity transmission system is proposed and examined. In a saturated color picture, picture quality was improved by the system as compared with a constant luminance system.

View full abstract

Subjective estimation on EDTV which has the high dynamic resolution characteristics is described. On a subjective 7-grade estimation scale, a picture quality improvement of 1-2 grade is obtained.

View full abstract

A three dimensional hole is introduced. It is existed in a horizontal, vertical and temporal high frequency region of alaminance signal. An additional horizontal detail signal is multiplexed in the hole.

View full abstract

This paper proposes a special EDTV in which the characteristics of the telecine images are being used efficiently. The Y/C separation, line-doubling, and the resolution of Y and C can be improved. The "Telecine-EDTV" has many merits in the transitional period of its introduction.

View full abstract

An advanced scan line conversion algorithm for letter-box processing is developed. Line converters of this system are based on progressive scan. A progressive scan signal is transmitted by an interlace scan signal and additional signals which are multiplexed in upper and lower black bars.

View full abstract

A scan conversion technique which uses a two dimensional (V-T) perfect reconstruction filter bank is described in this paper. An additional signal, which is low-pass filtered in the horizontal domain, is transmitted to the decoder. A compensation method which removes an artifact caused by the low-pass filtering is also described.

View full abstract

An intermediate type wide aspect EDTV that can be generated by digitized NTSC and HDTV system is proposed. In this method, letterbox processing and horizontal compression are used for wide aspect with some processings for picture quality. Then, we experiment these processings with hardware.

View full abstract

This paper describes experimental EDTV system with high quality wide aspect image and digital audio. In this system, letter-box method is used, a vertical-temporal augmentation signal transmitted in the top and bottom panels and a higher horizontal resolution component embedded on "Hole", and digital audio signals are sent by RF-QAM(Radio Frequency Quadrature Ampliture Modulation).

View full abstract

A widening method which uses a time-varying compression technique is described in this paper. Side-panel signals are up-sampled in the first place, then time-axis compressed and low-pass filtered. This technique enables to reject a time-varying low-pass filter. A method which improves the side-panel picture quality is also described.

View full abstract

A digital audio signal transmission system in which digital audio signals are multiplexed in the video carrier by using an quadrature modulation method is proposed. Ghost is generally in terrestrial broadcasting, then video signal appears as crosstalk in the digital audio channel. Two dimensional waveform equlizer and the line comb filtering reduced this crosstalk.

View full abstract

VSB filter and Nyquist filter which to be used for MUSE-VSB・AM modulation and demodulation have been developed by SAW filter having 12MHz band pass width at 58MHz band. Ripples of group delay of VSB filter are within 10nsec and of Nyquist filter are within 20nsec. Both of them have suitable characteristics for use in HDTV transmission on CATV.

View full abstract

A 2-dimensional filter LSI for MUSE decoder has been developed. The 1-D and 2-D filter block, which used to be composed of 5 LSIs on the former MUSE decoder, was integrated on a single chip. It has resulted in the reduction of LSI count and thereby, decoder's cost.

View full abstract

The two-dimensional interpolation filter with variable tap length for MUSE stationary domain has been developed. As the result of studying tap length of this filter, 5(Vertical) × 3(Horizontal) two-dimensional filter which was able to suppress aliasing interference sufficiently was obtained.

View full abstract

We examined the motion detector in the MUSE decoder which used interfield differential and slope mask techniques. We concl uded that the new methods were more effective than conventioanl motion detection methods.

View full abstract

This paper describes a programable clock driver LSI and its application to MUSE decoder, and a compact LSI MUSE decoder embeded with seven LSIs developed to reduce the size or expand the performance of the decoder. The board area of the new decoder is reduced to about 70% of the previous one.

View full abstract

This paper describes about Output Signal Processor LSI for MUSE decoder, which is successflly deveroped by NEC Corporation. The function of LSI is GBR-Matrix, enhancer, character inserter, and tre-levelSYNC generater using digital signal processing. The specification of LSI is 132-pin PGA package, 1.1Watts power, 1.2μm rule of CMOS process.

View full abstract

A single chip LSI has been developed to decode audio signals, that are multiplexed into MUSE singnals for transmitting the HDTV signals. The LSI requires just a few peripheral parts (such as two RAMs and a VCO circuit), use of the LSI can reduce the size of the MUSE audio signal decoding digital unit.

View full abstract

In order to receive the HDTV signal by conventional NTSC TV sets, a MUSE-NTSC converter which converts a MUSE-HDTV signal to a NTSC format signal is neccessary. Using a large scale integration by an advanced 0.8μm HC^2 MOS technology, the authors have developed two custom LSIs for a high-quality and low-cost MUSE-NTSC converter.

View full abstract