[in Japanese]

View full abstract

We have developed a small-sized camera module for mobile phones. In the modern world of sophisticates mobile phones customers expect to be able to connect to the Net and communicate with images. First, we set the specifications of the camera module to meet market demands. Then we developed a camera which met those demands. For mobile phones a low profile camera module with high image resolution and high sensitivity are indispensable. We developed a module with a height of only 4.6 mm with compact 3.3 mm optical length lens. It is a 170, 000 pixel PIA-CCD cameras for high image quality. We ascertained that the camera had achieved high sensitivity (less than 2 lx), high SNR, and high flame rate (20 fps).

View full abstract

We have developed optical-adjustment-free camera modules for mobile phones. Because of the currentboom in mobile phones with built-in cameras, easy camera assembly to speed unit output is important. Adjustment-free manufacturing helps camera module assembly makers deliver mass products with short turnovertime. In particular, focus-adjustment-free cameras can be made smaller and task time speeded up. We have examined the trade off between resolution and optical focal depth for mobile phones. And we have developed a new structure for camera modules that has an optical system set on the CMOS sensor to minimize the margin of error in assembly.

View full abstract

We present a high density single carrier 1024 QAM transmission scheme for digital cable TV systems. A transmission of approximately 50 Mbps within a single 6 MHz channel and with a higher spectrum efficiency than a conventional digital cable modem, is achieved. For evaluating an equalizer performance in a cable TV channel, we carry out computer simulations before hardware implementation of a 1024 QAM receiver. The BER result of a prototype 1024 QAM modem without forward error correction is 10^-4at a CN ratio of 37 dB with Gaussian noise. The implementation loss is less than 1 dB. In addition, we experimentally investigate robustness against interference and the influence of nonlinear amplification of a 1024 QAM signal.

View full abstract

Optical near-field and localized plasmon excited around metal particles was reported to have played an important role in the readout mechanism of a light scattering type super-resolution near-field structure (LSSuper-RENS) disk. However, neither of these has yet been proven using experimental data. We clarified that thermal effects induced by an incident laser light could work mainly on the readout mechanism in the LS-Super-RENS. Additionally, a super resolutional readout was also observed on the optical disk with different materials and structures, even though there was no metal particles. This meant that the optical near-field and localized plasmon were not contributed to the main readout mechanism. We proposed a new super resolutional readout mechanism based on Gaussian distribution of an incident laser light intensity, heat accumulation in a bubble pit shape, and thermoreflectance phenomena.

View full abstract

In OFDM signal reception, frequency dips caused by multipath propagation generate burst errors and this leads to a severe degradation of performance. To minimize this degradation, an error control code such as an error correcting code combined with a frequency interleaver is usually used. We analyzed the characteristics of a data spreading scheme in the frequency domain combined with an error correcting code. We ascertained that the proposed method results in a performance improvement. Moreover, we demonstrated that a reduction in the delay time and calculations in the decoder is possible.

View full abstract

A non-realtime modulation method, called time base modulation (TBM), has been studied and analyzed with computer simulations using fast fourier transforms (FFTs). Although TBM had been proposed and had the patents granted to the authors for constructing an analog modulation system with a bandwidth conversion technique for video signals, the characteristics had not yet been clarified. Using frequency analyses of correlations between sidebands of TBM and FM, in which TBM was not similar as it resembled FM, we confirmed that we could approximate frequency spectrum responses of TBM signals using Bessel functions with unique parameters different from modulation indexes of realtime modulation FM signals. The non-realtime modulation system can be applied not only to analog processing systems but also to new digital processing systems using digital coding modulation techniques with low noise characteristics similar to FM and time base conversion techniques of digital image information for high-speed transmittion systems.

View full abstract

The perceived speeds of gratings with various spatial frequencies, which shifted at the same speed, were evaluated using psychophysical methods. When the gratings shifted at low speeds, perceived speeds were found to be independent of the variation in the spatial frequencies of the gratings. However, at high speeds, an increase in spatial frequency was found to result in an increase in perceived speed. Furthermore, secondary polynomials were found to represent the relations between perceived speed and spatial frequency more accurately than the linear expressions claimed previously by other researchers.

View full abstract

We propose a surveillance system that is capable of obtaining a wide-angle view as well as detailed views of multiple regions of interest simultaneously. Based on a CMOS imaging technology, a solid-state multicamera-array whose pixels can be read out at random is used. The camera array does not have any mechanicalmotion elements and quickly switches between a low-resolution output for wide-area-monitoring and high-resolution output for detailed-views of multiple moving objects. Thus, the size of the total transmission data from the cameraarray to the monitoring-site can be kept relatively small. The system continuously outputs a hardware-subsampled wide-angle view. If moving objects enter the field of vision, the system automatically tracks them and outputs detailed views of all objects in addition to the wide-angle view. We have developed an experimental sized surveillance system with 16 CMOS random access image sensors.

View full abstract

We analyze the subjective perceived picture quality of coded stereoscopic pictures which are disparate in the object-and background-regions using a single stimulus evaluation method. In.addition, we examine subjective differences in perceptions of picture quality between stereoscopic pictures and 2-dimensional displayed pictures. When the picture quality in the object-and background-region is the same, no significant difference is seen in the evaluation value of the stereoscopic picture and the counterpart 2-dimensional displayed picture. When the picture quality of the object-region is different from the picture quality of the background-region, the obtained subjective picture quality of the stereoscopic picture is the same as or better than the evaluation value of the 2-dimensional displayed picture. The extent of this tendency correlates with the proportion of the object region to the entire image.

View full abstract

This paper discusses new methods for coding of multi-view images. Three distributed solutions are proposed based on a parent node and child node framework. A parent node encodes the whole image whereas a child node only partially. The proposed scheme allows independent encoding of each view. Experimental results show good performance of the proposed architectures at low bit-rates.

View full abstract

Watermarking has a trade-off relation between the number of embedded bits in a digital signature and image degradation. Therefore, we propose a watermarking method that improves the relationship between the number of embedded bits and image degradation. Experimental results show that image degradation using the proposed method is less than that with the conventional method when the number of embedded bits is the same.

View full abstract

We present a non-photorealistic rendering (NPR) method for generating pencil drawings and oil paintings from photographs. Strokes were generated using anisotropic smoothing filters (ASF). These filters were preliminarily reviewed and extended to NPR methods. In an NPR method for pencil drawings, the original photograph was transformed into a stippled image, which was then filtered by ASFs. Orientation of strokes was aligned with a gradient in luminance or with that of depth values. Cross-hatching type drawings and colored pencil drawings were also presented. In an NPR for oil paintings, random noises were added to the original photograph, which was then filtered by the ASF. Spatial weights in the filter were extended to bandpass characteristics for enhancing strokes. Two types of oil paintings were given. The first one used the gradient of the original photograph and another one used the gradient of the photograph after noise addition.

View full abstract

There are many computer graphics simulations such as movies, games, city reconstruction plannings and disaster drill simulations that involve human behavior. In almost all cases, however, human behaviors are pre-decided by manual because it is too difficult to generate human behaviors automatically. Therefore, we propose a psychology-theory based method that models human behaviors. Human behaviors are divided into two categories : personal behavior and interpersonal behavior. On the one hand, personal behaviors depend on people's schedules and minds, which are always affected by time, place and occasion. On the other hand, interpersonal behaviors depend on the relationships among persons they encounter. They vary depending on the relationships among the persons involved. We applied our psychology-theory based model to students in a virtual campus and demonstrated that the proposed method is a very useful way to generate human behaviors automatically.

View full abstract

In a cel animation workplace, non-photorealistic rendering (NPR) techniques are often applied to a 3D scene model, which is constructed by referring to the 2D hand-drawn images of the scene. There exist a lot of scene features that must be modeled from the 2D information. The accentuation effects are amongst the most importantfeatures in a hand-drawn scene. In this context, accentuation refers to adding hand-drawn shading of the scene. Thespecific region that an artist wishes the viewer to observe is drawn more brightly than the other regions. We propose a method of extracting the accentuation effects from a single 2D painted image of an animated scene. This method enables the automatic addition of the accentuation effects to each frame of the animation, observed from other viewpoints.

View full abstract

We present a of capturing the visual appearance of a real environment such as the interior of a room. We propose a method for generating arbitrary viewpoint images by constructing a light field with an omni-directional camera. In this method, the onmi-directional camera positions of input image sequences are automatically estimated by extending Zhang's homography-hased camera calibration method to omni-directional cameras. We also use a B-Tree data stnicture for the light field. to improve the efficiency of virtual view image synthesis. Thus our method allows the user to explore a virtual environment with a. wide field of view that achieves a realistic representation. To demonstrate the proposed method, we captured our lab's interior with an onlni-directional camera, and successfully generated arbitrary viewpoint images for a virtual tour of the lab environment.

View full abstract

In Web3D, a refined model representation requires a large number of polygons. It is difficult to transmit a large number of polygons on the Internet. It is necessary to compress large amounts of data using modelsimplification. The model data can then be transmitted efficiently. The receiver can obtain the transmitted data in a form as close to the original model as possible, using subdivision surfaces. We propose a new method of generating control polygons for piecewise smooth subdivision surfaces using an extended QEM as a model simplification. We make the vertices of the simplified model correspond to arbitrary points on the polygons of the original model, instead of making the vertices ofthe simplified model correspond to the points of the original model. This reduces computational time and minimizes errors between the original model and the subdivided model, when producing control polygons from the original model.

View full abstract

We propose a new method for synthesizing player viewpoint images from multiple view tennis videos.Our method uses two key techniques : virtual-view synthesis and player's viewpoint estimation. In the former, we divide the object tennis scene into sub-regions, which are the background, tennis court ground, players, and ball. For each sub-region, a virtualviewpoint image is synthesized by considering the geometrical conditions of each region. Then the virtual viewpoint images of the sub-regions are merged into a virtual viewpoint image of the whole scene. In virtual viewpoint image generation, view interpolation, which is a way of synthesizing images at an intermediate viewpoint to two views of the real object, restricts the viewpoint to a position between the two views. To avoid this restriction, we propose our first key technique, whereby we are able to flexibly position the viewpoint. In our second technique, the viewpoint is computed using epipolar geometry from the center of gravity of a player. By applying the computed player's viewpoint to the former technique, we can synthesize player viewpoint images. Experimental results demonstrate that the proposed method can successfully provide a video of the tennis player's viewpoint from multiple-view video images.

View full abstract

Some researchers have used inverse kinematics to generate human motions. However there are some applications that involve direct torque or force control, such as weight lifting and baseball hitting. Wedescribe a method of modifying motion-captured data by considering forces or torques as the sources of human motions.

View full abstract

A non-photorealistic rendering technique for generating colored paper-mosaic images is presented. The inputted photograph is firstly approximated using a mutually overlapped distribution of color disks. Each disk is then transformed to a torn fragment of color paper. Our method produces a paper-mosaic image composed ofvarious sizes of paper fragments.

View full abstract

Voronoi tessellations are widely used in an artistic filter plugged into photoretouch software fortransforming an input photograph into a stained-glass-like image. However, real stained glass is usually composed of smoothly curved lead lines instead of straight line segments in the Voronoi tessellations. In this paper, we present a method for transforming a photograph into such a realistic stained-glass-like image. Colors in the input image are quantized by using the median cut scheme and the boundaries of color segments are smoothed by quantized mode filter steps. Textures imitating frosted glass and bumps in glass near lead lines are attached by using an emboss filter.

View full abstract

Various virtual reality systems on the basis of 3-dimensional computer graphics have been researched. Applications involving two or more persons cooperating work were made for an experiment. In cooperative work in virtual space, communication and a high of reality are important. However, this research did not consider evaluating the operation accuracy of cooperative work in virtual space. Therefore, we built the functionto evaluate the operation accuracy of cooperative work in spherical displays, and we examined the operation accuracy using an experiment.

View full abstract