[in Japanese]

View full abstract

This paper proposes a new method of describing spatiotemporal images based on the concept of structured abstract index images. In spatiotemporal images, each region in an image sequence forms a tube. An original image sequence can be reconstructed by cutting out the sequence of cross sections from the group of tubes in these images. That is, a moving image sequence can be described compactly by the group of tubes. This paper first discusses the method of approximating static contours by Bezier curves. Approximation of surface tubes by Bezier surface is examined in detail. A problem with this method is that it does not work well if shape of the cross section changes considerably along the time axis. This problem can be solved by growing a new Bezier surface from passing points of Bezier curve and by combining the different Bezier surfaces.

View full abstract

An efficient video coding algorithm for coding the upper half of a human body is proposed. In the coding system, the facial area, which is a moving object in the image sequence, is extracted from both the sending and receiving sides without any control information. The candidate areas of the person are segmented using color information and the moving areas. The color information, which represents skin color, is extracted with a modified HSV color system. The moving areas are estimated using motion parameters of objects, which are utilized for constructing previous frames. The face is then extracted by merging the candidate areas based on hue elements of the modified HSV color system. The standard “Claire” and “Miss America” videos are used for the computer simulation, and facial areas are accurately extracted in all frames.

View full abstract

The conventional quantization control method, in which the buffer memory status is fed back to quantizer, is not the best method for very low bit-rate video applicable to picture communications on the ISDN or the PHS network. This is because it mainly depends on the number of bits previously spent and it causes the picture degradation. In this paper, a new video coding method with spatiotemporal control at pre-processor and adaptive quantization control at coding-engine is proposed. This method enables synthetic control, which consists of estimation of the picture characteristic before encoding and feedback of the number of bits spent by the picture. Details of the synthetic control method and the coding scheme are described, and the performance of this system is discussed.

View full abstract

This paper proposes an adaptive mode switching for error resilient video coding. The error resilient video coding system dynamically replaces reference pictures in inter-frame coding according to backward channel signaling. This system prevents temporal error propagation by preventing the encoder from using the erroneous picture as the reference picture. This system has two modes; the ACK mode is effective in more erroneous condition, and the NACK mode is effective in less erroneous condition. The mode switching mechanism and the error condition on the network is explained, and the performance of the system compared with the conventional H.263 system is shown by computer simulation.

View full abstract

A motion picture coding scheme with high error resiliency that is applicable to wireless communication is proposed. Two methods are introduced to minimize the degradation of decoded images : re-synchronization method recovers synchronization from bit-slip of a VLC and two-way decoding using reversible VLC. To recover erroneous decoded images, an adaptive refreshing method is applied in which the encoder selects non-stationary areas for refresh by using the coding mode of the past frames. Simulation results show that the proposed scheme has high error resiliency and is applicable to high error prone environments with BER larger than 0.1%.

View full abstract

Properly representing image structure is a major problem faced in developing a computer vision system. This paper proposes a concise descriptor called the shape operator for describing image structure. We first consider an image function as a surface, then establish the shape operator by utilizing surface curvatures. The shape operator can give rich, detailed descriptions in terms of topographical shapes and has invariance properties which are required in computer vision systems. We show usefulness of the shape operator with a texture classification application. Experimental results indicate that the shape operator gives a representation that is sufficiently detailed for characterizing texture images. In particular, it can satisfactorily mitigate the adverse effects of variations in lighting.

View full abstract

This paper reports the development of a new opticalflow detection method which encodes motion components in four image data samples taken by flow shooting. It detects opticalflows from the image data by a simple filtering and arithmetic process. Flow shooting is accomplished by low-speed transfer in vertical CCD registers and a multiple field shifts in IT-CCD image sensors. The method provides real-time opticalflow detection for simple signal processing. The effectiveness of the method has been verified in real-time shooting.

View full abstract

To recognize whole three-dimensional environments, it is necessary to obtain depth and height information, or three-dimensional information, of a scene. This paper presents an omnidirectional stereo method, which can obtain three-dimensional information simultaneously in all directions. The method first takes a pair of omnidirectional images at different heights using the omnidirectional imaging sensor, HyperOmni Vision, and then detects corresponding points between the pair of images, finally obtaining three-dimensional information simultaneously in all directions. Experimental results show the feasibility of the proposed method with real omnidirectional images. This paper also shows the computation of common horizontal binocular stereo images from a pair of omnidirectional images for virtual reality applications.

View full abstract

We investigate the ergonomic requirements for lightness, contrast and character size of reflective liquid crystal displays (LCDs) for young and old users. The 30 young subjects (mean age : 21.3) and 24 old subjects (mean age : 58.6) rated 25 lightness and contrast combinations for five character sizes displayed on hard copies that simulated reflective LCDs. The results indicate that the character size required is 15 point when the characters are viewing at a distance of 40 or 50 cm and an inclination angle of about 40 degrees. The required contrast ratio depends on the background lightness. When the background lightness L^* is 70, the required contrast ratio is about 4 for the 15-point characters. When the background lightness is 50, however, the contrast required is 6 or 7. These requirements are almost the same for young and old subjects. When the character size is smaller than 15 point, however, old subjects require very high lightness and contrast compared to young subjects. This difference may be due to age-related differences in the visual acuity for close targets.

View full abstract

A psychological experiment on inferring presence or absence of attention to individual visual stimuli presented in a brief exposure duration is performed. The experiment is based on both gaze point and the timing of a spontaneous blink. Each stimulus was randomly presented on either the left or right side of a display. Subjects are requested to appropriately allocate attention to both sides of the display. Although the presence of attention to only about 50% of all stimuli based on gaze point is inferred, the presence of attention to over 80% of all stimuli can be inferred by adding the blink timing together as a physiological measure. These results suggest that a spontaneous blink tends to occur immediately after the completion of detecting a peripherally presented stimulus.

View full abstract

Basic technologies of a drive system with digital drivers that have gray-scale interpolative functions are reported. The driver reported in this paper is a 3-bit driver, although it is the first digital driver in practical use that performs gray-scale interpolation. It has also become the starting point of the following higher multi-gray-scale drivers. The driver is designed by 5 volt withstanding process that is the standard process of a digital LSI. It should be used in combination with the AC common driving scheme (ACCDS). The ACCDS itself is not a new concept, however, the floating gate driving method proposed in this paper makes the ACCDS possible without deteriorating picture quality. In addition in this paper, the issue of the transparency asymmetry with respect to polarity of the drive voltage is discussed and the importance of estimating the asymmetry compensation errors inevitable in a drive system that performs gray-scale interpolation is discussed. It is also shown that the floating gate driving method can be simplified as long as the compensation errors are under a certain limit.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract