Reports of the Technical Conference of the Institute of Image Electronics Engineers of Japan
Reports of the 237th Technical Conference of the Institute of Image Electronics Engineers of Japan
Showing 1-39 articles out of 39 articles from the selected issue
Date: March 7- 8, 2008 Location: Hiroshima University
Friday, March 7
10:00-12:05 Chair: Fumitaka ONO, Tokyo Polytechnic University
  • KokSheik WONG, Kiyoshi TANAKA
    Session ID: 07-05-01
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    This paper proposes a novel data hiding method in the MPEG domain in which the image quality of the modified video is completely preserved to that of the original (compressed) video. To our best knowledge, there is no data hiding method that completely preserves the video quality after data embedding, and this method is the first attempt of its kind. Mquant value and quantized DCT coefficients in a macroblock are carefully manipulated to embed information while completely preserving the image quality. This method is reversible where the modifications done during data embedding could be undone to restore the original video. This method is applicable not only to existing MPEG1/2 encoded videos but also to the encoding process of MPEG video from a sequence of raw pictures. The problem of filesize increase as a result of data embedding is addressed, and three independent solutions are presented to suppress the filesize increase while trading off with payload and coding efficiency. Basic performance of this method is verified through experiments on various existing MPEG1 encoded videos.
    Download PDF (384K)
  • Eri TOJO, Hernan AGUIRRE, Kiyoshi TANAKA
    Session ID: 07-05-02
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    A watermark sharing scheme that generates high quality halftone images using Genetic Algorithm (GA) was extended to the method that decodes NC2 kinds of different watermark images from any two images selected from N generated ones. Although this method can clearly decode watermark images consisting of bold letters like Logo, it becomes difficult to detect watermark images consisting of fine letters or texture. In this work, we modify the evaluation function on watermark appearance so that we can decode fine watermark images in this method. We verify the performance of this method through computer simulation using some benchmark images.
    Download PDF (1844K)
  • Yasuo TAKEHISA, Kiyoshi TANAKA
    Session ID: 07-05-03
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    An improved super-resolution method that achieves dense motion estimation (DME) with multiple motion vectors using overlapped block matching (OBM) was proposed. Also, a computational cost reduction method that uses the statistics of multiple motion vectors was proposed under the condition that preserves the original performance of the improved super-resolution method completely. In this work, we investigate the possibility to achieve further computational cost reduction and image quality improvement simultaneously when the constraint that completely preserves the performance of the improved super-resolution method is relaxed. Experimental results show that we can further reduce computational cost up to around 20.7 - 27.8% while slightly improving or keeping the quality of reconstructed images depending on observed images.
    Download PDF (646K)
  • Naoki OTA, Hernan AGUIRRE, Kiyoshi TANAKA
    Session ID: 07-05-04
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this work, we investigate color halftone image generation using Genetic Algorithm (GA). So far, some halftoning methods using GA for monochrome image were proposed, which generate high quality halftone images with precise gray-levels and appropriate contrast. However, simple combination of R, G, B halftone planes independently generated by this approach may cause interference noise among three planes. To avoid this problem, in this work we propose a color halftone image generation method that evaluates individuals using Yuv signals transformed from RGB signals for population evolution of GA.We verify the superiority of this method through objective and subjective evaluations for generated halftone images.
    Download PDF (2784K)
  • Keita TAKEUCHI, Kiyoshi TANAKA
    Session ID: 07-05-05
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Recently, monitoring systems using surveillance camera and/or web camera are frequently used in many applications. Captured videos are often compressed and stored with MPEG format, which is the international coding standard for moving pictures. Among several MPEG series, MPEG-4 is expected to be more utilized by implementing it on portable devices such as mobile phone. From this point of view, in this work, we try to detect moving objects directly on the compressed video sequence by MPEG-4. The object detection is conducted by MB (macro block) by MB by considering kind and amplitude of motion vectors, and relationship with neighbor MBs. With the method developed in this work, we could detect about 80% of the region of moving objects that we want to extract while somehow containing unnecessary regions.
    Download PDF (830K)
13:00-15:05 Chair: Kiyoshi TANAKA, Shinshu University
  • Michiaki TOYODA, Kazuhisa OBA, Shin'ichi NAKAMURA, Tetsuya YANAGIMOTO, ...
    Session ID: 07-05-06
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Class marks defined by the Nippon Decimal Classification are written on book labels and used for arrangement of books. Books shelved incorrectly are difficult to find, so librarians walk around between bookshelves and check correctness of books' location. This work called “ shelf reading ” is a time-consuming job. Our goal is to construct a support system for shelf reading in libraries. We proposed a holistic matching method of handwritten book class marks toward finding books in incorrect bookshelves. In this paper, we consider the utility of this method for supporting shelf reading from a result of experiments.
    Download PDF (305K)
  • Nobuhisa TAKANO, Madoka HASEGAWA, Shigeo KATO
    Session ID: 07-05-07
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    JPEG2000 is an international standard for still images coding. MQ-coder, which is a type of arithmetic coder, is used in JPEG2000. The calculation algorithm of the sub-interval in MQ-coder is simplified and then the operation cost is reduced. However, the coding efficiency is decreased. Therefore, we studied binary arithmetic coding method for improvement of the encoding efficiency. In the conventional methods, the sub-interval sizes are calculated based on several symbols in a queue as exactly as possible. As a result, the encoding efficiency of the conventional method increased about two percent than that of JPEG2000. In this paper, we propose more efficient method than conventional one by improving the probability estimation. Simulation results show that the encoding efficiency of the proposed method is more efficient than that of the conventional method.
    Download PDF (450K)
  • Takashi MASUDA, Nakoyuki AIKAWA
    Session ID: 07-05-08
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    The JPEG standard is one of the most popular image compression schemes today.However, a major drawback of JPEG is that the compressed image exhibit visually annoying blocking, mosquito noise and ringing artifacts, especially at high commpression ratio. Mosquito noise is typically seen when a sharp edge separating two uniform regions occurs within a block.In this paper, we propose a method of reducing distortion of near edge by making small steep change of the luminosity ingredient leading to a mosquito noise. Moreover, we ease deterioration of visual quality by adding compensation to the image using the theory of projections onto convex sets (POCS).Computer simulatios show that the proposed method could achieve a smaller file size compared with some other methods.This approach is easy to implement and is fully compatible with the existing JPEG standard and enables reaal-time application.
    Download PDF (425K)
  • Jun YOSHIKAWA, Makoto KOSUGI, Nobuhiko MUKAI
    Session ID: 07-05-09
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Extraction of Meta-data from Sumo video and making its digest is investigated. First, a scene structure of a Sumo match is defined, each scene is then distinguished based on features of image and characters such as color and size of distinctive area. As a result, meta-data of 98% on real battle scene is extracted, and also that of 99% on win/loss and winning trick, 100% on wrestler's name are extracted. Additionally, it becomes possible to make a digest of 3 to 9 minutes from one hour sumo's video and to make a digest of specified wrestler or winning trick.
    Download PDF (2062K)
  • - Comparison of Multi-Level Image based Method and Black Number based Method -
    Fumitaka ONO, Makoto KITAMURA
    Session ID: 07-05-10
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In order to find an efficient lossy coding scheme of ordered dither images, two different coding categories, multi-level lossy image coding (LML) with post-processing of ordered dithering, and sending the number of black pixels in a block of bi-level pixels (BPN), are proposed and studied. Among the former category, we adopted JPEG2000 as the best one, and for the latter category, we assumed the context based coding, which refers to the numbers of black pixels in several neighboring blocks. By setting same bit rate per pixel for these two methods, the picture quality of the restored lossy bi-level images were evaluated using the PSNR of inverse halftoned images. It was found that to use the inverse halftoned (IH) image of lossless bi-level one as the reference image in calculating the PSNR will be better than to use the original multi-level image in reflecting the subjective picture quality, and that BPN has better performance than LML for higher bit rate and LML will catch up BPN as the bit rate decreases.
    Download PDF (348K)
15:25-16:15 Chair: Kazufumi KANEDA, Hiroshima University
16:15-17:30 Chair: Kazufumi KANEDA, Hiroshima University
  • Mei KODAMA
    Session ID: 07-05-12
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Recently, we had often used many video and music contents based on video coding standards over the broadband internet. However, there are some problems for contents distribution systems. Then, DRM technologies are used for contents service in broadband, network and storage media. For example, access control methods, quality transform schemes and encryption method for media or network are studied. We propose a content delivery system based on video data structure. It consists essentially of an media transform method using structure divider and combination processor, and an authentication method of contents and users. In this paper, from the viewpoint of data transmission, we consider the feasibility of our system.
    Download PDF (719K)
  • Yuuki TERASAWA, Kazumi KOMIYA
    Session ID: 07-05-13
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    A wide-ranging fire detection technology in outdoor has been required. For this requirement, ordinary detection systems were studied based on complex shape change of smoke. However, in addition to this feature we focused on change of transparency with the time passage. Then, We proposed an new extraction technique of smoke from features such as density, direction of the movement and shape. Concretely, based on difference between frames, amount of feature and judgment algorithm were clarified by experiment From the results, the validity of the proposed algorithm were confirmed and some study subjects such as accuracy improvement and application into fire monitoring system. were picked up for advancement.
    Download PDF (1702K)
  • Junya ITOH, Akihiro NAKAMURA, Jong Geol. PARK, Kotaro MATSUSHITA, Kenn ...
    Session ID: 07-05-14
    Published: 2008
    Released: October 07, 2008
    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 is being developed. Using the system's feature which allows the search for similar image characteristics, forest and field fire area retrieval was tested as an application of the image data analysis system. A GUI frontend has been added to the system, and performance improvements and additional functionality to allow Web support is currently being developed. This paper gives an overview of the satellite image data analysis system, describes the improved features, and reports on the performance evaluation results.
    Download PDF (2554K)
Saturday, March 8
9:00-10:40 Chair: Mei KODAMA, Hiroshima University
9:00-10:40 Chair: Hidekazu SEKIZAWA, Toshiba Tec Document Processing Systems Co.,Ltd.
  • Tatsuhito SATO, Kota AOKI, Hiroshi NAGAHASHI
    Session ID: 07-05-19
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this paper, we propose a new method for extracting the human airway area from three dimensional chest X-ray CT images. Extraction of human airway from chest X-ray CT images is critical importance for both the computer aided detection of lung cancer and the virtual endoscopy system. This algorithm is composed of the following 2 steps; 1.Generating computatoinal region:2.The airway segmentation based on 3-D mathematical morphology: Steps 1 and 2 are performed iteratively.
    Download PDF (1303K)
  • Kyohei HOSONO, Kota AOKI, Hiroshi NAGAHASHI
    Session ID: 07-05-20
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this paper, we propose an animation system in which a non-rigid agent constructed by mesh structure autonomously learns more effective motions from past actions and generates some actions to reach to a given goal. A spring-mass sysyem is implemented to represent flexible geometry deformations of the non-rigid agent. In our system, forces and torques are provided due to mesh deformations of the agent, and their resulting motions are simulated. Several control points and the goal of the agent is given by a user. A selected parameter is updated by both crossover and mutation techniques. The agent repeats the deformation of its shape and aims at the goal.
    Download PDF (1105K)
  • Shuichi ASAJIMA, Hiroshi NAGAHASHI, Kouta AOKI
    Session ID: 07-05-21
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    This paper describes a 3D reconstruction system from image sequences which is acquired by a single camera.This system consists concretely of the following three steps: 1) 3D reconstruction of the features in the image sequences by factorization, 2) estimation of the camera parameter by RANSAC and integration of 3D data, 3) integration of high resolution data acquired from zoom function into low resolution data.Based on this system, user can get the 3D information of the scene and the detailed texture information.
    Download PDF (4810K)
  • Kohei TAKAMURA, Isao MASUIKE, Toru TAMAKI, Kazufumi KANEDA
    Session ID: 07-05-22
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    To render translucent materials realistically, it is necessary to calculate subsurface scattering, which is light scattering inside the materials. In this paper we propose a method for simulating subsurface scattering physically, where the subsurface scattering equation is discretized with using a beam of light. In addition we try to render a material with two layers. Our goal is rendering a multilayer translucent material.
    Download PDF (521K)
10:55-12:35 Chair: Yushi KOMACHI, Osaka Institute of Technology
  • Shota NISHIMURA, Miki MATSUMOTO, Osamu Uchida
    Session ID: 07-05-23
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In recent years, glasses fitting systems and hairstyle simulation systems by using face images taken by cellular phones have been developed. However, in many of them we have to correct the face position manually and it is inconvenient. In this study, we consider eye's position detection method for face images taken by cellular phones. The method for eye's position detection by using SSR filter has been proposed, which is inadaptable for large variation of face size, though. In this study, we propose a fast and high accuracy eye's position detection method using multiple SSR filters and two new filters called Eye filter and TSR filter.
    Download PDF (433K)
  • Miki MATSUMOTO, Osamu UCHIDA, Shohachiro NAKANISHI
    Session ID: 07-05-24
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Face detection plays an important role in various applications, such as face recognition, video surveillance and human-computer interface. In the present paper, we propose a new face detection method using the probability maps. The basic concept of the proposed method involves the use of the property of the deviation of the position of the complex region and flat region of the face. Experimental results indicate that the proposed method is more robust with respect to variation of scale and rotation and partially interruption compared to conventional method.
    Download PDF (289K)
  • Manabu SHINOHARA, Yuta GOTO, Osamu UCHIDA
    Session ID: 07-05-25
    Published: 2008
    Released: October 07, 2008
    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. It is required for digital watermarking method that the watermark is perceptually invisible and robust against various kinds of attacks. However, it is known that digital watermark is weak to geometric transformation in general. In this paper, we propose an improved hybrid digital watermarking method which robust against geometric transformation. Moreover, we realized more robust watermarking method by the matching of the SIFT features.
    Download PDF (350K)
  • Hiroyuki YAMADA, Ryousuke FUJII, Mitsuyoshi SUZUKI, Hiroshi ITO, Shige ...
    Session ID: 07-05-26
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    The paper presents a new digital watermarking system where the watermarks survive after re-shooting a screen. The system consists of 2 essential components, a watermarking device and a detection software. The watermarking device embeds watermarks into HD moving pictures and outputs watermarked pictures in real-time. The detection software captures moving pictures, analyzes them, and extracts data which are embedded as watermarks from them. We verified that the detection software can extract 35-bit watermark data from 5-minute re-shot moving pictures.
    Download PDF (821K)
10:55-12:35 Chair: Yoshinori HATORI, Tokyo Institute of Technology
  • Kazunori MIYATA, Kaisei SAKURAI, Toshihiro TOMOI, Hiroshi TASHIMO, Koj ...
    Session ID: 07-05-27
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Southern Ishikawa area research unit, which has been selected as one of the promotion of science and technology on city area project, is developing a visual simulation method for traditional craft materials. This paper reports a method for representing gold leaves and Japanese lacquer faithfully using Spectral Bidirectional Reflectance Distribution Function (S-BRDF). This paper also describes about a method for laying out craft materials and for generating tearing patterns of gold leaves.
    Download PDF (2753K)
  • Tomoya MIWA, Junichi MATSUDA, Yoshiki MIZUKAMI, Katsumi TADAMURA
    Session ID: 07-05-28
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    We propose a method for watermarking authentication information into a printed matter taking into account the appearance as a document with a dot pattern code based on an illustration. Since random dots are usually used for conventional dot pattern, a document with that dot pattern often gives readers strange feeling. On the contrary, we avoid appearing this kind of problems by using an image which is obtained through transforming an illustration into a dot pattern image as a dot pattern code. We developed two key techniques; one is a method for transforming from an illustration image to a dot pattern image which is suited for watermarking; the other one is a method for watermarking that allow us to reduce noise caused by errors in printing and scanning a document. We verified availability of the proposed method through an experiment on perception of subjects by appearance of printed images with a dot pattern code made by the proposed method.
    Download PDF (1255K)
  • Rie KINOSHITA, Yoshiki MIZUKAMI, Katsumi TADAMURA
    Session ID: 07-05-29
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    We propose a method for rendering a realistic computer generated landscape with falling snow. The proposed method consists of three elements: a movement model of a snowflakes considering with the effect of local wind; efficient and accurate collision detection between snowflakes and objects relating with snow accumulation; simple snow accumulation model. The falling trajectory of a snowflake is complex because the snowflake's figure is complicate and the trajectory is deeply influenced by the wind having temporal-spatial localities consequently. We made some simplifications relating to figure and movement of a snowflake for expressing realistic snowfall with practical computation cost. We also propose a new method for collision detection between snowflakes and objects using multiple depth maps, which are corresponding to multiple view directions. A simple model of snow accumulation is developed for expressing a transient situation on a surface of an object at beginning of snow. We implemented the proposed method and confirmed the usefulness of it through some snowfall animations.
    Download PDF (1930K)
  • Kazuya WAKIMOTO, Daisuke MEGA, Nagaharu HAMADA
    Session ID: 07-05-30
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    A town navigation service on a mobile phone etc. has been increasing in recent years. The CPS is mainly used in such position estimation terminals. However, in a room of a bad electric wave conditions, it is difficult to use. In this report, we propose a estimation method of camera position from a one-shot image of indoor, and the validity of the proposed method has been shown through experiments.
    Download PDF (373K)
13:25-15:30 Chair: Mutsuo SANO, Osaka Institute of Technology
  • Kakeru KITAJIMA, Keiji SHIBATA, Yuukou HORITA
    Session ID: 07-05-31
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    It is not possible to take a picture of a general digital camera because the shutter speed and the exposure are not sufficient in the place where the illuminance looks like nighttime and a dark place. In this time, therefore, only low dynamic range luminance image can be obtained. To improve the dynamic range for lower luminance part, we used the infrared image information. By using the infrared image information, we improve the dynamic range of the visible image.
    Download PDF (2167K)
  • Shin-ichiro YOSHIDA, Masahumi HOSONO, Daisuke IWATA, Keiji SHIBATA, Yu ...
    Session ID: 07-05-32
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    The road status estimation technology from road surface is one of the important issue as ITS (Intelligent Transport System) in recent years. In this paper, we present road status estimation method from road surface image using texture analysis and neural network. By using this method, the road risk information (puddle, snow, ice and etc) from the status of the road surface can be obtained. Overall estimation accuracy for all mesh images is achieving over 90%.
    Download PDF (1272K)
  • Yuji HAYASHI, Yuta MATSUMOTO, Keiji SHIBATA, Yuukou HORITA, Kazunori H ...
    Session ID: 07-05-33
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    The method to detect wet condition from road images using polarization feature have been proposed. In the method, we have to use two cameras for taking horizontal and vertical polarizing images at a same time. In this paper, we propose omni-directional polarization image procedure method using omni-directional camera and polarization filter. This method enables to take horizontal and vertical polarizing images at same time.
    Download PDF (3360K)
  • Kei MAEDA, Keiji SHIBATA, Kouji ONISHI, Yuukou HORITA, Masato TAJIMA, ...
    Session ID: 07-05-34
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    The construction of the SNOW hazard map is very important with the same as the hazard map of the EARTHQUAKE and the TSUNAMI. In this paper, we investigate the snow hazard map, and construct the prototype that used Ajax. Weather information was made as XML database, and the layer of the hazard map was obtained by the spatial interpolation method. By using these database and layer informations, a scalable, real-time hazard map was constructed in the different ground resolution.
    Download PDF (1560K)
  • Yoshikazu KAWAYOKE, Teruhiko KAIODO, Keiji SHIBATA, Yuukou HORITA
    Session ID: 07-05-35
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this research, we have analyzed the influence of transmission quantity of video and sound on realistic sensation in order to develop the objective multimedia quality assessment model. Therefore, we have conducted the subjective assessment by the Semantic Differential (SD) method, and analized the factor of the subjective assessment result. As a result, three kind of factors were obtained. It was guessed that the three kind of factors are the overall quality of video and sound, the quality of video, and about the silence of contents. Moreover, the realistic sensations obtained by adjective pair of “stereognostic sense” was able to be estimated with high accuracy from score of the three kind of factors. Furthermore, the three factor scores were able to estimate from the transmission quantity of video and sound. Therefore, it turned out that transmission quantity of video and sound influence the realistic sensations.
    Download PDF (1879K)
15:50-17:30 Chair: Tetsuro KUGE, NHK Science & Technical Research Laboratories
  • Kazuya IKEMURA, Mutsuo SANO
    Session ID: 07-05-36
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    A gesture plays an important role in emotional communication between human and robot. But conventional motion capture methods require special equipments and special markers to detect a locus of human's motion. In addition, the lack of data occurs frequently. In this paper, we focused on tracks of face and both hands based on Particle Filter. The likelihoods of face and hands are calculated based on distance and flesh color features, and applied to particle filter. Acutually, we reliably detected the locuses of nine representative emotional motions.
    Download PDF (285K)
  • Yuki IZUMI, Mutsuo SANO, Satoshi NISHIGUCHI
    Session ID: 07-05-37
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In order to realize the smart communication of human and robot, it is necessary for the systems to observe human's behaviors and recognize their intention and attension. Eye-Movement and motion of face play a major role to recognize human's intention and attention. This paper estimates the visual direction using iris contours, which are detected from an input image and fitted with ellipses. We use particle filter algorithm for tracking the iris of the eye stably against blinking and propose a likelihood function based on a brightness of iris region. In addition, we detect nodding and discuss the components of nodding.
    Download PDF (385K)
  • Manabu SASAKI, Ayumi SAKITANI, Yushi KOMACHI
    Session ID: 07-05-38
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    When multimedia information is rendered and presented, pets feel a strong interest in the information, and their attention will be caught by the multimedia equipment. Therefore, multimedia equipment should be more significantly designed than other equipment, and some design guidelines are required. IEC/TC100 discussed those requirements and suggested that a technical report should be developed. Responding the requirements, this paper studies some behavior of cats to multimedia information and proposes a design guideline for multimedia equipment.
    Download PDF (1871K)
  • Toshiya SUZUKI, Yushi KOMACHI
    Session ID: 07-05-39
    Published: 2008
    Released: October 07, 2008
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    TrueType font file is one of the most popular scalable font resource used in computer related environment. In 2007, ISO/IEC JTC1/SC29 had adopted the Microsoft enhanced version “OpenType” as the standard font resource to be embedded in MPEG-4 system stream, and OpenType specification version 1.4 was standardized as a part of MPEG-4 standard, ISO/IEC 14496-22 Open Font Format (OFF). ISO/IEC 9541, the existing standard to interchange the information of font resource had been designed to be independent with specific file format and character encoding schemes. On the other hand, OFF is bound to Unicode encoding system and a few file formats that are specified out of ISO/IEC 14496-22. In this report, we compare the positions of ISO/IEC 9541 and 14496-22, and summarize the issues that is caused by the technical gaps between two standards. Finally, we report the working of ISO/IEC JTC1/SC34 to update ISO/IEC 9541 for capability to deal OFF.
    Download PDF (1350K)
feedback
Top