Reports of the Technical Conference of the Institute of Image Electronics Engineers of Japan
Reports of the 243rd Technical Conference of the Institute of Image Electronics Engineers of Japan
Showing 1-32 articles out of 32 articles from the selected issue
Date: March 5- 6, 2009 Location: Kagoshima University
Thursday, March 5
12:30-14:25 Chair: Tsutomu HORIKOSHI, NTT DoCoMo, Inc.
  • Yuta GOTO, Osamu UCHIDA
    Session ID: 08-05-01
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In recent years, there have been many studies of digital watermarking as one of the way to protect the copyrights of digital content. Some digital watermarking methods tolerant to digital-to-analog conversion have been considered, and moreover navigation systems from analog information to digital information for camera-equipped cell-phone have been proposed recently. In this paper, we proposed a digital watermarking method for printed materials embedding the watermark into hue component. The results of experiments of extracting the watermark from the printed materials by using the scanner indicate that the proposed method is tolerant to digital-to-analog conversion.
    Download PDF (477K)
  • Nobuhiro NOJIMA, Makoto KOSUGI, Nobuhiko MUKAI
    Session ID: 08-05-02
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    This research proposes a selection method of image features that are effective for searching landscape images on content-based image retrieval. In the method, all the images are divided into square blocks, such as 16 x 16, 32 x 32, and so on. Next, the resemblance between a query image and ones in the database, is calculated based on the similarity for the corresponding blocks of them. The features of color, edge, wavelet and fractal, are selected from among many kinds of image features by a preparatory test. As the result of the experiment with the proposed method, more than 80% images retrieved with the selected features have satisfied user's subjectives.
    Download PDF (1665K)
  • - Parallel computing for inspections and image processing -
    Hiroyuki KOKUBUN, Yuki CHAZONO
    Session ID: 08-05-03
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    CPU Load for image processing increases rapidly due to computational complexity to handle huge amount of data captured by high-resolution color camera. Therefore we investigated GPGPU to decrease the process time. The process time is evaluated for some image inspection processes, such as binarization, shading correction, look-up tables, color inspection, and filters. As a result, the use of GPGPU improves the performance of image processing.
    Download PDF (317K)
  • Yuko OUCHI, Ikuko UWANO, Masashi KAMEDA
    Session ID: 08-05-04
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    It is important to discover a brain disease at an early stage to decrease the mortality. However, the present medical examination for the brain needs large expense of both money and time. In our study, we have proposed a new MRI device to find out any brain diseases even if people don't come to hospital. The output of the proposed device is a set of super-low field MR images which are indistinct in comparison with the previous MR images. Therefore, it is difficult to detect brain diseases from the super-low field MR images directly. This paper presents a method to detect brain tumors and brain bleeding from the super-low field brain MR images of patients. Our method can be specified the position of disease based on the following properties; i ) the left and the right brain is almost mirror image each other, and the correlation between them is higher, ii ) the intensity characteristics including disease in the super-low field brain MR images are changed gradually like contour lines. In the experimental results, it is shown that the correlation coefficients for the diseases the left and right brain tend to be lower compared with the other candidates.
    Download PDF (543K)
  • Junya ITOH, Akihiro NAKAMURA, Jong Geol. PARK, Kotaro MATSUSHITA, Kenn ...
    Session ID: 08-05-05
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    As part of the Academic Frontier Project at Tokyo University of Information Sciences, a system which analyzes and retrieves requested image data from the database of collected MODIS satellite image data has been developed. In this research, the current system was modified to support web access. A web server and application server was introduced to the server configuration, and the client application was ported to a web browser based GUI. This paper gives an overview of the Web-based satellite image data analysis system, and describes the performance evaluation results.
    Download PDF (2850K)
14:35-15:20 Chair: Hiroshi OCHI, Nihon University
15:30-17:02 Chair: Naoki KOBAYASHI, Saitama Medical University
  • Ryo KURIHARA, Tetsuri INOUE
    Session ID: 08-05-07
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Our experimental study examined the effect of the foreground stationary pictures on the strength of vection. We used a CAVE-like box type multi-screen 3-D display. Onset velocity, which is the velocity of the moving pattern when the subject feels vection , was compared among the foreground images: no foreground image, a frame image, the inside of a train, and the inside of a room. From the experimental results, a virtual scene simulating outside scenery of a moving object may induce a strong sense of vection and so enhance the sense of reality.
    Download PDF (367K)
  • Miki KONAGAYA, Yoshiki MIZUKAMI, Ken MATSUDA, Katsumi TADAMURA
    Session ID: 08-05-08
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this paper, we propose the system which generates automatically the illumination pattern which took a user's liking into consideration from the musical piece described in SMF format. The production lighting which utilizes programmable LED lighting, such as illuminations of Christmas, and aims at the synergistic effect with a musical piece is getting popular. If we can provide the means which makes the illumination pattern for each musical piece without needing special technical knowledge, it is expectable that an application field spreads. In the proposed method, we pay attention to tonality, tempo, pitch, and volume of a musical piece as the important items. These items are extracted from a musical piece described with SMF format. We then define a rule which translates these four musical items into timing of a lighting change and color and light volume at that time. The illumination pattern is generated based on user's liking and the above rule. We evaluated the sense of unity between a musical piece and the illumination pattern which was generated by the proposed method through the subject experiment. The experiment was conducted by playing back the musical piece and lighting of LED lighting apparatus simultaneously.
    Download PDF (468K)
  • Ryota NAKAMURA, Shigeru AKAMATSU
    Session ID: 08-05-09
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    This paper describes an automatic vision-based spoken word recognition system that utilizes, instead of audio signal, visual motion signal which is obtained from motion pictures taken of a region around the mouth during speech. Motion information on each pixel in the input time-series imagery was obtained by computation of optical flow, and feature values representing a spatial configuration of pixel-wise velocities were extracted for each frame image. Both starting and ending points of time for each spoken word were defined using the velocity feature values, and a high dimensional feature vector was obtained to indicate time variation of the velocity distribution within the period of utterance. As a preliminary performance evaluation of the proposed feature in spoken word recognition, discrimination test of five spoken words including A-RI-GA-TO-U and KO-N-NI-CHI-WA was conducted, and fairly promising results were achieved.
    Download PDF (449K)
  • - Discrimination between Human and Portrait by Eyewink -
    Magoichiro SUZUKI, Naoki SHIBA, Shigeru AKAMATSU
    Session ID: 08-05-10
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Recently, varieties of biometric authentication systems were investigated and some of them came into practical use. Among them, face authentication has the advantage of being socially-acceptable as well as less demanding to users, because humans usually take the same measure to person identification in daily life. The system, however, should respond to a challenge made by "Masquerade" through the use of photographic portraits. Thus, a new scheme for preventing such type of masquerade is proposed in this paper. Since true human could hardly repress eyewinks, the basic idea is to detect them under the framework of template matching. What is characteristic about the proposed scheme is to prepare templates not only for the eye regions but also for other undeformable region of the face, such as around the nose. This would enable to distinguish the case when the eyes were lost of tracking because of rapid motion of the face from the case when true blinks were detected. When even a single blink has not been detected during a certain time period, the input face image is judged as a photographic portrait. Validity of the scheme for preventing masquerade in face authentication system was tested by experiments and fairly promising results were obtained.
    Download PDF (324K)
17:12-18:52 Chair: Yoshinori HATORI, Tokyo Institute of Technology
  • Yohei MOMOKI, Kohta AOKI, Hiroshi NAGAHASHI
    Session ID: 08-05-11
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    A mobile robot may associate landmark objects with their spatial and structual attributes in its internal map to recognize 3D environment. However, the landmark registration by hand according to the environment usually takes much cost. Therefore, this paper proposes a method for image based landmark registration by using SIFT features. In the method, images taken by a digital camera are associated with an internal 3D map created by the robot.
    Download PDF (5306K)
  • Kiminori SATO, Takahiro OONO, Masayuki KASHIMA, Mutsumi WATANABE
    Session ID: 08-05-12
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    At present, in making to be a code number by bank ATM, the numeral of 4 digits is mainly used. On the other hand, card falsification and unjust photographing in the code number input don't decrease. Then, the contrivance that the numeral arrangement of input screen of the code number is changed has also done it. And, the bank which establishes the vein authentication system using palm and finger which entered the practical stage recently increases. However, though the technique according to the biometric information of the vein certification is the high security, since card falsification and unjust photographing are difficult, on the other hand the introduction cost of it becomes considerable large amount of money. And, though it is the biometric information of fingerprint and vein of hand, there is a case in which it can not be used by the failure. And, the certification which contacts the equipment can not deny either the psychological resistance. In this study, security key entry system using infrared camera is developed. This proposal system allocates four shape patterns of hands as a code number, and it is made to be the security key.
    Download PDF (1022K)
  • Makoto HIROSE
    Session ID: 08-05-13
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this paper, I contrive the system that can be effectively used various range images by various industries and different field by web cooperation working on web. This paper discribes a integration module of range images on this system. In order to integrate range images by coorperation web working, we need to regist range images of different size, position, aspect and resolution that offers different institute, company and general user. We try to regist by method of semiautomated registraion formed 2-staged formula between visual integration by hand and auto integration using extended gaussian sphere of range image. Also, in order to edit and view range images on web, we need a stress-free operation and browsing. However, increasing in the showing polygon number are slowed down showing speed, and it is made editing and viewing range images slowly. So, we need to hold down polygon numbers of range image using editing and viewing works. And so, we devise a method to modify division number of view campus relative to them. In this paper, we provide efficacy and try to improve the slow viewing.
    Download PDF (461K)
  • Koji KUMA, Yasuhito TAKEUCHI
    Session ID: 08-05-14
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Download PDF (1009K)
Friday, March 6
8:30-9:39 Chair:Mitsuji MATSUMOTO, Waseda University
  • - On an Effect of Adaptive Selection of Interpolation Method Guided by Maximum A Posteriori Estimation -
    Tetsuro KUGE
    Session ID: 08-05-15
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    A method to improve the image resolution by the Wavelet Analysis and the directional interpolation on the spatial domain is now under development. The method conjectures more densely located new samples which are spawned by the directional two dimensional interpolation utilizing the two dimensional shift-variant filters. In this report, an effect of the adaptive selection of the interpolation measure guided by the Maximum A Posteriori Estimation on the correlativity between the pixel value to be conjectured and the feature statistic of the neighboring pixel regions.
    Download PDF (109K)
  • Ami YOKOYAMA, Akira KUBOTA, Yoshinori HATORI, Shigeyuki SAKAZAWA
    Session ID: 08-05-16
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this paper, we present a novel method for reconstructing a super-resolution image using multi-view low-resolution images captured for depth varying scene without requiring complex analysis such as depth estimation and feature matching. The proposed method is based on the iterative back projection technique that is extended to the 3D volume domain (i.e., space + depth), unlike the conventional superresolution methods that handle only 2D translation among captured images.
    Download PDF (1211K)
  • Yuta TAMURA, Yasuo TAKEHISA, Kiyoshi TANAKA
    Session ID: 08-05-17
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In fractal image coding, we first divide the entire image into non-overlapping range blocks. Then, we approximate each range block using the contraction map of a domain block that is extracted from the same image with a larger size than range block. Since the method utilizes selfsimilarity contained in the image to be coded, the fractal parameters (IFS codes) extracted can be directly used for image enlargement. However, the quality of enlarged images using IFS codes is not always sufficient due to unpleasant noise caused by block approximation. To solve this problem, in this work, we propose three methods to improve the quality of enlarged images using IFS codes, i.e., (i) the method using multiple domain blocks (MDB), (ii) the method using multiple range blocks (MRB), and (iii) the method combined (i) and (ii) (MRB+MDB). We verify the performance of the proposed methods through computer simulation.
    Download PDF (1050K)
9:39-10:48 Chair: Tsutomu HORIKOSHI, NTT DoCoMo, Inc.
  • Kohei SAITO, Issei MATSUNAGA, Minoru ITO
    Session ID: 08-05-18
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Precise detections of scaling factors and scale-change center positions for input images is needed for correcting scale-changed input-images. In this paper, scale factors for images with different scale-change center positions are detected by SPSA technique. Furthermore, a new simple and precise detection approach for analytically detecting scale-change center positions is proposed. From conscientious experiments using three kinds of image types, it becomes clear that scaling factors of input images can be detected accurately independently of scale-change center positions and that the center positions can be located with as a high accuracy as averagely 1 pixel except for a few cases. These results show the effectiveness of the proposed approach.
    Download PDF (403K)
  • Yasuo TAKEHISA, Koji HARA, Kiyoshi TANAKA
    Session ID: 08-05-19
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this work, we investigate a temporal interpolation method which accurately reconstructs frames that are not included in the original video to increase the temporal resolution of the video sequence. In the interpolation precess, two (forward and backward) motion vectors are obtained between two existing frames by block matching, and are used to estimate motion vectors on the frame to be reconstructed with the precision of sub-pixel. Our approach generates high-precision (high-quality) frames by using these motion vectors as they are with the precision of sub-pixel without rounding them off to the precision of pixel (integer). Through experiments we can see that our method improves the PSNR of the reconsrtucted frames about 3[dB] compared with the average interpolation method with the precision of interger. Also, the problem that the performance deterioration as we increase the interval between frames is addressed, and a countermeasure using overlapping block matching is investigated.
    Download PDF (638K)
  • Jakchai BUTSRIKUI, Mitsuji MATSUMOTO
    Session ID: 08-05-20
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Abstract All of current On-line handwriting recognition text input method for Thai scripts systems is based on traditional writing system or natural writing. They tried to fulfill user satisfaction by providing a user friendly and easy-to-use system. Those systems allow user to write freestyle unconstrained handwriting, which is perceived as very convenient. Nevertheless, in practice users have found an error in recognition frequently. Our proposed method aims to reduce the error rate through a different way of thinking. We design an easy-to-write constructed scripts based on a dot line of the scripts, which children use for their handwriting practice. We will not be concerned about headings and lines, but points. Only significant points which represent that script will be used in recognition process. If we draw the straight line between those points, the shape of the script will be similar to one of the simple unconstrained handwriting, which is easy to write. We hope that with this method the error recognition rate will be reduced.
    Download PDF (2267K)
10:58-12:30 Chair: Naoki KOBAYASHI, Saitama Medical University
  • Kazuhiro MANABE, Sei IKEDA, Masataka IMURA, Yoshitsugu MANABE, Kunihir ...
    Session ID: 08-05-21
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this paper, we propose a luminance correction method of underwater images with consideration of light absorption and scattering. In this method, we use light attenuation model calculating an observed pixel value from reflectance of the object surface, depth, lighting condition, scattering coefficient and absorption coefficient. At first, we measure a reflectance known marker. We optimize scattering coefficient and absorption coefficient. Next, we correct an underwater image from the image of the measurement target and optimized coefficients. The effectiveness of the proposed method is verified through experiments.
    Download PDF (8947K)
  • Natsuki OGAWA, Koichi KAWABATA, Koshi OSADA, Minoru ITO
    Session ID: 08-05-22
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In conventional camera calibrations, the camera parameters are calculated so that the image of each mark becomes equal to an image synthetically generated with the expected camera parameters. This procedure is based on the indirect approach. This paper describes a new camera calibration technique based on the direct calibration approach, in which the camera parameters are directly and visually obtained from the trajectory of the optical ray incident to each sensor chip in the camera. It is made clear that the parameters can be calibrated with a high accuracy and the proposed technique is highly effective for the camera calibration.
    Download PDF (473K)
  • Daisuke MEGA, Nagaharu HAMADA
    Session ID: 08-05-23
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    As progressing of introduction of a multi-axis machining tool, in order to avoid the collision of the tool head and a working object, estimation method of the 3-dimensional shape of the object becomes very important. In this paper, we are proposing the idea of 3-D shape estimation method using the tool marks on a surface of the working metal, and report that we can remove the shape distortion from a reflected tool marks by processing of few pictures of surface image of working object, and the possibility of 3-D shape estimation is obtained.
    Download PDF (409K)
  • Takuya YOSHIBA, Kenji UENO, Minoru ITO
    Session ID: 08-05-24
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    This paper proposes a new technique for passively detecting surface positions and shapes of glass plates. Though there are lots of approaches such as the stereopsis in computer vision, those methods are hardly effective for detecting transparent objects such as glass plates. Recently we have proposed the technique for detecting mirror surfaces, in which the rays of entering and reflecting at the surface are found and reflecting positions are detected geometrically with the triangulation principal. However, it could not be applied to detection of transparent objects such as glass plates, because of reflection rays from rear surfaces. In the proposed method, a new type of the pattern displayed on the pattern plate is used for discriminating rays from front and rear surfaces of glass plates. From the experiment, it is made clear that the proposed technique is highly effective for detecting front surfaces of glass plates.
    Download PDF (2376K)
13:20-14:52 Chair: Fumitaka ONO, Tokyo Polytechnic University
  • Yasuhito TAKEUCHI
    Session ID: 08-05-25
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    An innovated photogeny (beauty of photo) is proposed using specific/selected wavelength band intensity for luminance in combination with visible (normal eye-sight) color (hue) information. When near-infrared band intensity is colored by visible color, the product image has both infrared band transparency (visibility) for distance and natural or more innovated color impression. When R, G, or B images are each colored by visible color, they make each characteristic unique image for vision impression. Application with suggested photogeny is underway.
    Download PDF (937K)
  • Yosuke SHIMADA, Toru TAMAKI, Kazufumi KANEDA
    Session ID: 08-05-26
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    High dynamic range imaging (HDRI) has become a common practice in computer graphics providing more accurate light measurements, thus allowing more convincing rendering results. Typically HDR data is stored as RGB intensities but this is not suitable to simulate several optical and perceptual phenomena in the real world, such as interference phenomenon and diffraction phenomenon. Such phenomena require spectral light distribution information to be properly simulated. However storing spectral data can be expensive so compact ways to represent such data type are desirable. This work presents a novel HDR image data format to compactly store spectral data by combining two different orthonormal basis functions: cosine function and wavelet function.
    Download PDF (446K)
  • - Changing hue due to overall Luminance condition -
    Michihiro MIKAMO, Yousuke SHIMADA, Marcos SLOMP, Toru TAMAKI, Kazufumi ...
    Session ID: 08-05-27
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    High dynamic range images (HDRI) have been employed in computer graphics to enhance visual quality since they provide more accurate real world measurements of light. Displaying HDRI in typical displays, however, requires tone mapping as the dynamic range between the image and the device can differ significantly. Although several algorithms were proposed, most of them do not account for proper human vision light response, such as mesopic vision. In short, mesopic vision relates how hue changes due the overall scene's luminance: blue-shift occurs in low-light conditions, making red intensities appear dim. This work proposes a novel tone reproduction technique to simulate mesopic vision using spectral-based HDR images.
    Download PDF (538K)
  • Yusuke UCHIDA, Hiroyuki KASAGAMI, Makoto KOSUGI, Nobuhiko MUKAI
    Session ID: 08-05-28
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    This paper proposes a method that can reproduce a 3D landscape with altitude by combining 2D map data with Laser Scanner (LS) data. LS data has altitude data, which are measured by a laser irradiation device equipped with an airplane. However, in the region where woods or rivers are located, LS data can't measure the accurate altitude. Then, the exact altitude data around the river should be extracted by removing the altitude data of buildings or woods from the LS data. Next, the LS data which reach the riverbed should be distinguished to estimate the elevation of the riverbed. On the other hand, the river area can be extracted from 2D map. And, the river contours are represented by curved lines. With these results, 3D scene of the river and its surrounding area can be represented with exact altitude by combining the altitude data of the ground surface and the riverbed.
    Download PDF (1237K)
15:02-16:34 : Sakuichi OHTSUKA, Kagoshima University
  • Yusuke MORIZUMI, Kenji MATSUO, Akira KUBOTA, Yoshinori HATORI
    Session ID: 08-05-29
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    We propose a new authentication system using video which is taken during moving handheld camera in front of the face. The proposed system uses the parametric eigenspace method to extract both individuality and motion features from the obtained image sequences. The proposed method employs the DP-matching to calculate the track similarity. We conducted verification to show the effectiveness of proposal. The experimental results show that continuous tracks plotted on the multidimensional eigenspace contain motion features. And by the experiment verifying the accuracy, we achieved EER=3.2% .
    Download PDF (1815K)
  • Yuta ASHIKAWA, Noriyoshi OKAMOTO
    Session ID: 08-05-30
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this paper, we proposed a human detection method under local brightness change. In recently year's buildings taking in outdoor light aimed for an illumination cost cut. Therefore an outdoor local brightness change occurs in the indoor and becomes the obstruction of the human detection. We improve a robust background subtraction method Radial Reach Correlation (RRC) for local brightness change. RRC maintains robustness for local brightness change by using texture from correlation between pixels pairs. This method copes with a local brightness change by creating the texture of the fixed domain. The experimental result show the possibility of detection by the daytime local brightness change and the night general brightness change.
    Download PDF (308K)
  • Shun FUKUMOTO, Masayuki KASHIMA, Kiminori SATO, Mutsumi WATANABE
    Session ID: 08-05-31
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Today, operation to system is intentional. There is no non-intentional one. If the system action without made to person consider about the system, it can do a more natural and harmonized support. For that, the system needs recognition his condition. The purpose of this study is to presume internal state of person who is in animation. Now, we assumed the target state warm and cold . It presume his inner state by recognizing behavior that seems appear at the state.
    Download PDF (192K)
  • Takashi HOSHI, Noriyoshi OKAMOTO, Kiyohiko ABE
    Session ID: 08-05-32
    Published: 2009
    Released: November 06, 2009
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In recently years, The research to detect the human eyes direction to input a character, etc. into a computer is performed. This research is advantage which it needs only eye movement and the people who was confined exercise function. Many eyes input system that have been devised by the present is many things with special device. There is the eyes input system which can use it under natural light with home video camera as a precedent research of this research. However, this system is necessary what supporting item for the head don't move greatly. Therefore this report proposes to correction of greatly head movement which the system for a heavy orthopedically impaired is to make more general.
    Download PDF (382K)
feedback
Top