This paper proposes a method for estimating three-dimensional (3D) motion parameters of a rigid object, of which surface is covered with a texture, by three images from a single camera. The method estimates unit normal vectors of texels and the 3D shape of the rigid object in each image by using a Shape-from-Texture technique. The rotation parameters are obtained from average vectors of the normal vectors at the three images. The translation parameters are determined as the displacement between the centroids of successive shapes. Based on the 3D motion parameters, a correspondence between texels in successive images is established. The correspondence can provide a more exact shape of the object through stereo triangulation. The estimation of exact shape and 3D motion parameters should prove to be useful technique for the content-based image retrieval.

抄録全体を表示

As reflected by the launch of satellite digital broadcasting in Japan at the end of 2000, the trend toward more advanced multimedia-oriented broadcast programs is gaining momentum. To deal with this trend and to support program production work in an efficient manner, we previously proposed a hierarchical program-information management structure that can be used throughout the program-production process, from planning to recording, editing, and secondary use. Using this structure, we have developed a hierarchical program-production technique and used it to construct a prototype system. The new technique was tested at a program-production site and was found to be efficient in terms of managing and reusing program information, and highly useful in providing a comprehensive view of story development. This paper describes the proposed technique and its features and presents an overview of the prototype production system.

抄録全体を表示

The study on image-based method, such as image-based rendering, image-based object, promotes a new problem called Image-based Shadow Generation. Image-based shadow generation aims at generating new shadows from a few reference images. As a start of the image-based shadow generation problem, building, as a representative of man-made object, is chosen as our study object in this paper. To solve the problem of image-based building shadow generation, we propose a new technique called shape from shadow silhouette, which reconstructs 3D model of building. This technique consists of three steps : shadow extracting & transforming, shadow complementing, and shape modeling. By this technique, the rough 3D model of building can be recovered from several reference shadow images, then new shadow can be generated easily and quickly by this recovered model. The experimental results are given in this paper. The efficiency of our approach is demonstrated by the results.

抄録全体を表示

人物頭部が動いたり, 顔表情が変化しても顔を抽出するため, 色ヒストグラムインタセクションの類似尺度を用いて物体を抽出する手法を改良し, 顔の抽出に適用した.改良では, 従来のアクティブ探索法をより高速化する手法を考案した.本手法を用いて, 顔抽出実験を行い, 顔抽出の安定性と処理速度に関する評価を行った.

抄録全体を表示

本研究速報では、MPEG符号化されたサッカー映像を対象として、映像信号に対する復号処理を行うことなしに、符号化データを直接用いてシーンの自動分類を行う方法について述べる。特に、MPEG符号化データに基づいて、シーンの分類のために利用できるどのような情報がどの程度の割合で検出できるかについて報告する.

抄録全体を表示

We introduce a temporal-spatial picture model, analyze the spectra of pictures in flat displays, and discuss the effects of spatial or temporal specifications of pictures on picture quality. We found that reducing the inter-pixel gap every two pixels to increase brightness or resolution results in uneven pixel pitch. An appropriate rate of unevenness produces higher brightness or resolution and improves overall picture quality. We also found that the use of a field dispersed PWM, which disperses the driving pulse to increase the apparent field frequency, repress the flicker that often occurs with high brightness or large size displays. Moreover, we found that using a TV camera to shoot pictures for the display of a field dispersed PWM results in good picture quality and enables image signals of different field frequencies (such as NTSC or PAL) to be displayed normally. These findings are important steps toward the goal of achieving high-quality tiled displays.

抄録全体を表示

For the purpose of the improving luminous efficiency of PDP, a cathode featuring a low discharge voltage has been developed employing a thin film of Al electrode covered with Y₂O₃. This structure enables the cathode to have long life. By applying the Y₂O₃/A1 cathode, the DC-PDP can be operated using 100 V less than an ordinary panel can be using an Al cathode. In comparison to MgO as a protecting layer of AC-PDP, the Y₂O₃ has excellent secondary electron emission characteristics. In order to clarify the conduction mechanism, which is necessary for a dc discharge, the surface and depth profile of the cathode is examined with SEM, RBS, and AES. Many small cracks are found in the Y₂O₃, which are caused by the high-temperature process. From the cracks, an Al cathode bus line is sputtered during the aging process and the surface of the Y₂O₃ is covered with a thin island-like Al film and a conduction path is formed through the cracks.

抄録全体を表示

We previously developed a field-sequential stereoscopic display system with a 42-inch diagonal HDTV DC-type plasma display panel (PDP). This system uses an LCD shutter stereoscopic eyeglass, synchronized with the left and right pictures on the display by infrared signals. Stereoscopic images were displayed on the PDP using the proposed light emission scheme and gray-scale signal processing. However, it was found that cross-talk impairment was incurred due to leakage between light for the left and right images. We calculated the amount of cross-talk as a function of phosphor persistence and decay time and estimated the desirable decay time for a field-sequential stereoscopic display system. This enabled us to develop a signal processing method for reducing the cross-talk impairment, and a subjective evaluation confirms the effectiveness of the method.

抄録全体を表示

In video processing, compression techniques are required to improve the efficiency of transmission and storage of video data. Data compression methods can be classified into two basically different categories : lossy compression methods and lossless compression methods. Lossless compression methods are required for medical image processing and for compressions broadcast materials. We developed a novel video lossless compression method using the Hilbert curve, by mapping the three-dimensional data to one-dimensional data along the curve and then applying adaptive linear prediction coding to the one-dimensional data. Experimental results show that better compression ratio is obtained with our method than with lossless JPEG, although the processing time for both methods is almost equal.

抄録全体を表示

The purpose of this paper is to clarify the effects of visual information on corresponding auditory information when they are presented simultaneously in a wide frontal area (180 degrees) on a horizontal plain. Sound localization experiments were carried out both with and without visual stimuli associated with sound stimuli. In the experiments, four different visual images of a human speaker and the four corresponding speech sounds were used as the visual and sound stimuli, respectively. The visual stimuli were projected on a half-cylindrical shaped screen in five directions (-60, -30, 0, 30, 60 degrees), and the corresponding sound stimuli were reproduced in 13 directions (-90, …, -15, 0, 15, …, 90 degrees) via loudspeakers placed in each direction or transaural reproduction using a pair of loudspeakers (±30 degrees). The degree of sound localization accuracy was evaluated according to a five-grade system in cases where visual stimuli were absent. The results obtained showed that a sound image far away from its corresponding visual image (maximum 150 degrees) was affected by the visual image, and that a sound image having a relatively low degree of localization accuracy tended to be strongly affected by the visual image. These results show that, in addition to the relative position of visual and sound stimuli, the degree of sound localization accuracy is one of the important elements in auditory-visual interaction on a wide frontal area.

抄録全体を表示

In this paper, we propose narrow element spacing arrays consisting of level sensitive phase multipliers (LSPM). A level sensitive phase multiplier comprises down converters, frequency multiplies with a phase locked loop (PLL), and an amplitude modulator.
In order to confirm the capability of these arrays, we first examine the receiving patterns of two element arrays using the half wavelength dipoles and three element Yagi-Uda antennas. The number of the multiple function “n” is set to ”2” for measurement. The experimental results show that the quarter wavelength spacing arrays with LSPM have the equivalent receiving pattern of the half wavelength spacing arrays without LSPM. Second, the four element arrays using inverted-F antennas with LSPM, which have a 0.2 wavelength spacing, yield the equivalent pattern of a 0.8 wavelength spacing arrays without LSPM. In this case, the multiple function “n” is set to“4”.

抄録全体を表示

音響再生装置の伝達周波数特性 (振幅・群遅延周波数特性) の平坦化を行い, その再生音を評価した.特に高度感性情報再現に重点をおいた.その結果, 従来オーディオ評価語に関する音質は改善されたが, 高度感性情報に関する音質は劣化した.その原因追求の実験の結果, 特性補正のために用いたDSP装置に起因する物理要因が高度感性情報再現を劣化させているに違いないことが明らかになった.すなわち, 高度感性情報再現のためには, 音響再生装置の伝達周波数特性平坦化以前に考慮すべき重要な物理要因が存在していることが明らかになった.そのひとつは再生されたアナログ信号の中に観測される時間方向の伸び縮みひずみ (jitter) であることを確認した.これまでの研究もふまえて検討すると, 想像をはるかに越える小量のjitterでさえも高度感性情報再現を損ねると考えられる.

抄録全体を表示

本論文では, NHK技研のフルカラーインテグラルフォトグラフィシステムで撮影された画像を入力として, さまざまな視点位置における画像をインタラクティブに合成する手法を提案する.提案手法を実装したシステムで実験を行い, 画質向上の有効性を検討した.

抄録全体を表示

液晶ディスプレイは, CRTディスプレイと比較して立上り, 立下り時間が遅く, また発光が1フレーム期間持続的であるために動画像を表示した際の画質低下がいわれている.しかし, 実際の液晶ディスプレイにおいて動画質を評価しCRTディスプレイと比較検討した報告はない.そこで, 本研究ではTN, IPS, STN各液晶モードの液晶ディスプレイ上に表示された動画の画質をCRTディスプレイと主観的に比較評価した.評価パターンとして黒地に白の線がその場で回転するものと, 白黒のチェッカパターンが水平方向に移動するものを使用した.前者は視線移動が少ない場合を想定し, 後者は視線が対象を追従して動く場合を想定した.また, 液晶ディスプレイの応答速度が階調レベルに依存することを考慮し, 白と黒の階調の組合せを0を最小レベル, 100を最大レベルとして, 0-100, 0-50, 50-100, 50-75の4種類とした.主観評価結果によるとSTNモード液晶ディスプレイは他のモードの液晶ディスプレイに比べて明らかに画質低下が認められたが, TNモード, IPSモードでは, いくつかの条件を除けば画質差は見られなかった.CRTディスプレイとの画質差が検知されない限界速度は, 線パターンでは, 519～1038deg/s (TN, IPS), 130～230deg/s (STN) となった.また, チェッカパターンでは, TNモードの0-100, 0-50条件およびIPSモードの50-75条件では, 168～350dots/s, TN, IPSモードの他の条件は, 78～168dots/s, STNモードは78dots/s以下となった.
エッジぼけ幅を基にCRTディスプレイとの画質差が検知されないための立下り時間を検討したところ, 1～2msが必要であることがわかった.

抄録全体を表示