Reports of the 269th Technical Conference of the Institute of Image Electronics Engineers of Japan
-
Taichi KUDO, Takafumi SAITO
Session ID: 13-04-0
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We consider visualization techniques for analysis of large-scale time-series data. Purpose of analyzing time-series data is to find the characteristics of them, and to predict the variation of them. In time-series data analysis, it is common to find factors in variation of the data. However, it needs detailed knowledge of the data. Therefore, we propose some visualization techniques to find special behaviors of time-series data, and factors of them. We analyze electric power consumption data of TAT, find variation factors of it, and visualize them with calendar-based visualization technique. Then, we can overview whole data, and find trends and special behavior of it.
View full abstract
-
Naoki KANESHIGE, Yoshiki MISZUKAMI, Akira OHGAI, Shinji IKARUGA, Takes ...
Session ID: 13-04-02
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Improvement in quality of an evacuation exercise can be expected by using a simulator which constructs an urban area in the virtual space and allows us to have an experience of evacuation exercise in the virtual area. The above simulator needs to render not only destroyed building but also rubble on the road in the virtual space, in order for the simulator to provide correct choices for decision making of evacuation to the user. In this paper, we propose a distribution model of rubble which takes into account the degree of collapse of the building due to an earthquake. We confirmed the usefulness of the proposed method through visualizing the distribution of rubble obtained by the proposed method.
View full abstract
-
Tsuyoshi HARADA, Michihiro MIKAMO, Bisser RAYTCHEV, Toru TAMAKI, Kazuf ...
Session ID: 13-04-03
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we propose a system for displaying high dynamic range spectral (HDRS) images using local tone mapping. HDRS images store light intensity according to the wavelength of light using more than 8 bits for each voxel. Since HDRS images are much larger than conventional RGB images, they require higher computational cost for displaying. The proposed system operates in two stages; 1) compressing and recording an HDRS image, and 2) uncompressing and displaying it. In the displaying process, we use a local tone mapping based on bilateral filtering. The system can display an HDRS image interactively with various parameter settings using a graphical user interface.
View full abstract
-
Msafumi KOSUGI, Osamu UCHIDA
Session ID: 13-04-04
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
An effective combination of music and video images generates efficacious interaction between auditory and visual processing. Then, these techniques are used at concert halls, on promotional videos, or on TV commercials. However, the professional knowledge and/or tools are necessary to create video images considering the interaction between auditory and visual processing. From this kind of circumstances, we have proposed a method to create a video image well-suited to the music based on its characteristics and chorus section detection automatically. However, there is an issue that the accuracy of chorus section detection was not high. Then, in this study we tried to review the chorus section detection method and change the shape and motion of objects. The results of the verification experiment show the effectiveness of the method.
View full abstract
-
Tomoyuki NISHITA
Session ID: 13-04-05
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In computer graphics field, the geometric model of Utah teapot is a famous standard primitive. This is represented geometric model including curved surfaces, which is called as Newell teapot. The model is used for evaluation examples on research papers, and the CG images of this teapot are introduced on many text books. The author also used computer generated Utah teapot in our papers. Recently I have visited the Computer history museum in Silicon Valley. The original teapot and its CG images are displayed as a contributed model to the evolution of computers and CG technologies'. I recognized the contribution of the Utah teapot again, so I surveyed the papers which used Utah teapot, and would like to report the contribution of the Utah teapot to CG evolution.
View full abstract
-
Tetsuya Ito, Masashi Baba, Shinsaku Hiura
Session ID: 13-04-06
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Images that have some viewpoints are sometimes used in Animation or Art. Such images are called multi-perspective image, and are rendered by two approaches. One approach is ray tracing using combined rays. Another approach is deformaing objects for orthogonal projection. But,the images rendered by two approach are not its same. So, we present an approach that objects are deformed by FFD in order to be same orthogonal projection and ray tracing are the same.
View full abstract
-
Navamani Isaivathi, Sho Ikeda, Bisser Raytchev, Toru Tamaki, Kazufumi ...
Session ID: 13-04-07
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Spectrum-based rendering uses spectral distributions instead of just three RGB color components for representation of light sources and surface properties in the rendering equation. Since spectrum has a value at every visible wavelength, the spectrum-based rendering gives much accurate color computation compared to the RGB-based rendering and it also gives us an opportunity to simulate wavelength dependent phenomena and effects caused by spectrum difference. In this paper, we compare the relations between a spectral rendering and a RGB rendering using measured spectral intensities of light sources and measured spectral reflectances of objects. By this way, we investigate the importance and the performance of the spectral-based rendering.
View full abstract
-
Miki NAKASHIMA, Yoshiki MIZUKAMI, Katsumi TADAMURA
Session ID: 13-04-08
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The online market expands by the popularization of the Internet year by year. The products include many variations, and garments comprise a large percentage among them. In comparison with the purchase at the store, the advantage of the online purchase is that customers can choose a favorite one from a wider range of garments. However, the problems are that they cannot get an impression in wearing the garment because it is not possible to try it on before purchasing and that they cannot confirm the coordinates with their own garments. This study focuses on the second problem and proposes a virtual fitting system in which photo images are efficiently translated to the garment model and a physics engine, PhysX, is employed for realistically rendering the garment coordinates.
View full abstract
-
Shogo ABE, Takafumi SAITO, Toshihiro KOMMA
Session ID: 13-04-09
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We propose a making motion data method for compute generated cel animation. 3DCG can use motion capture data. However, Motion capture data cannot be used as it is. Because, japanese cel animations use a unique representation. We solved by the use of superlatives this problem. First, we give exaggerated to increase the operation using the Laplacian of Gaussian (LoG). By using LoG, the angle of the joints is changed appearance. In addition, We increase the display time of a key frame and create an exaggerated animation. By exaggerate the motion data in this way, We obtain a motion data suitable for cel animation.
View full abstract
-
Makoto GOMYO, Yoshiki MIZUKAMI, Katsumi TADAMURA
Session ID: 13-04-10
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we propose a system to send information of disaster reduction efficiently and watch over the people requiring assistance at a disaster. Disaster reduction for the people requiring assistance at a disaster, and the isolation reduction and health management for them are the most important administrative issue for the local government. The proposed system provides the watching function of everyday to reduce isolation of the people requiring assistance at a disaster, in addition to the safety confirmation function in the event of disaster. The watching function aims to promote the mobile terminal daily use of the people requiring assistance at a disaster. This function realizes both the maintenance of operating status and the practice of a terminal handling. Therefore, the transmission of information and the confirmation of safety can be carried out quickly and reliably when a disaster happens. We had developed a prototype of the proposed system and confirmed that the estimated function was realized.
View full abstract
-
Ryo YANAGAWA, Yoshiki MIZUKAMI, Katsumi TADAMURA
Session ID: 13-04-11
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Storm and flood damage such as the typhoon and a localized torrential downpour tend to become a larger scale than heretofore because of recently abnormal weather.Measures utilizing information Technology (IT) against a large earthquake and a tsunami have been prepared.However, many local governments with small budgets have not enough done the preparation with IT measures against the storm and flood damage.We propose the prototype-system to solve above problems.More specifically, we propose simple system which allows local government to gather the disaster related information easily and to accumulate and share the information.
View full abstract
-
Yutaro AOKI, Osamu UCHIDA, Kazumi TERADA, Yuki SATO, Kiyonari FUKUE, K ...
Session ID: 13-04-12
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The Great East Japan Earthquake that occurred on March 11, 2011 brought serious damage in the Tohoku district in Japan. There have been many examples of utilization of satellite image analysis to grasp the situation of damage. However, henceforth, it is more important to monitor how the destructed environment will recover for a long time. From such a background, our research group is working on a project such as “A project on the recovery monitoring of the destructed environment in the disaster area by satellite data analysis and field work and the practice of earth environmental education.” The main purposes of the project are to monitor the status of the environment recovery from both aspects of satellite image analysis and field work with high school students in the disaster area, and to practice earth environmental education for young people through such work. Moreover, to construct Web-based system to present the status of the environment recovery and the reconstruction of the disaster area clearly and to transmit this information all over the world are also purposes of the project. In this paper, the part of the progress of the project, especially the topic of the prototype construction of the Web-based system to present the status of the earthquake disaster recovery and the system of automatic classification Image data getting by field work are described.
View full abstract
-
Masahiro ISHIKAWA, Manabu KAWABE, Hideki KOMAGATA, Takashi KANO, Naoki ...
Session ID: 13-04-13
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Dialysis fluid in hemodialysis is administered to a patient’s body via inner filtration effect and replenisher in hemodialyzer. Therefore, a lot of dialysis fluid that is made in facilities usually needs to keep high cleanliness on the level of not having negative effect on patients even if administered to the patients. Viable cell count method is generally conducted by plate culture and membrane filter methods in order to check the cleanliness. Unlike agar media, the membrane filter does not have transparent medium so cannot extract with simple binalization. Moreover, the filter may have lattices so have problems to identify the lattices as colonies if simple methods are applied. This research is a technology to easily, accurately and automatically count the number of colonies cultivated on the membrane filter.
View full abstract
-
Takayoshi SUKETOMO, Yoshiki MIZUKAMI, Katsumi TADAMURA
Session ID: 13-04-14
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we reconstruct the three-dimensional structure by stereo algorithm using image segmentation. After a reference image is segmented into small regions according to the color distribution, the disparities of the regions are optimized based on a cost function. In particular, in order to prevent the disparity integration of regions on different planes, the closeness coefficients are determined based on strength and confidence of edges between the regions and the neighboring regions, and then incorporated into the cost function. The simulation results clarified the effect of the closeness coefficients for the disparity accuracy.
View full abstract
-
Sho FUJIHARA, Yoshiki MIZUKAMI, Akira OHGAI, Shinji IKARUGA, Takashi K ...
Session ID: 13-04-15
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The evacuation exercise simulator based on the walk-through system is suitable to provide a pseudo experience of evacuation.However,it may need time to master the operation of the movement in the virtual space and this drawback may become a barrier when we use this simulator as a tool for evacuation exercise for ordinary people.In this paper,we propose a method for automatic generation of a passable route for the walk-through taking into account obstacles.Specifically,the proposed method realizes automatic generation of the route which allows us to avoid the collision against the obstacles assuming the rubble on the road and to move the virtual space with easy operation.
View full abstract
-
Seira HAYAMA, Masafumi KOSUGI, Shun KITAHARA, Hiroki OGAWA, Yutaro AOK ...
Session ID: 13-04-16
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In recent years, the spread of digital camera, the evolution of camera mounted on mobile terminal, and the increasing the capacity of storage media are progressing. According to a certain survey, more than 70% of travelers said that they take photographs actively, and then the number of the photographs owned by each user is increasing rapidly. However, it is often difficult to manage these photos by the user itself because the number of varieties of subject of photos on a journey is large in general. In this paper, we propose an image classification method by considering the characteristics of tourist photographs to facilitate the management the photos taken at tourist spots. In this study, we propose a method to classify photos into three categories, “person”, “landscape”, and “food”, because the number of tourist photographs belonging to these categories is a lot in general. The results of the verification experiment using 230 actual photos shows the effectiveness of the proposed method.
View full abstract
-
Mei KODAMA
Session ID: 13-04-17
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Due to the popularization of video content services over the network, it is expected to use a lot of contents more. Since the improvement of efficiency in video content delivery is needed, the video content delivery methods using caching had been studied. Moreover, it is one of significant issues to manage and provide video data of multiple qualities in order to realize that the users use the video contents in various information terminals. Then, we had studied the video caching and delivery methods with scalable data. In this study, a service model of content utilization is defined. Priority calculation rules are proposed in cached contents management using hierarchical data for multiple qualities. q is defined as high content access ratio. By the simulation experiments, in the case of q = 0.6, proposed method is maximum about 11.4[%] superior to previous method in terms of TR , and in the case of q = 0.8, it is maximum about 17.6[%] superior. Therefore, we can especially consider that propose method is more effective when q is higher than 0.5.
View full abstract
-
Haufeng YU, Mei KODAMA
Session ID: 13-04-18
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The video contents accessed from different device have been increased. As a multiple quality solution, scalable coding has been studied. The common issue of these scalable coding is coding efficiency degradation due to layered related information. To solve it in this study, we propose representative MQ linear prediction method(Method B, Correction Method B' of Method B) and local MQ linear regression prediction method(Method C). MQs in every layer are encoded independently in the past. In these methods, reducing the amount of higher layer MQ information can be achieved by using the linear prediction from the lower layer MQ information. To evaluate the effectiveness of these, by simulation experiments, it was confirmed that the linear prediction can reduce layered information. In Method B, the information reduction effect by linear prediction was confirmed. Method B' improved the predict accuracy of Method B by using correction process. Method C is possible to improve the prediction efficiency in Method B and Method B' by using local MQ to predict directly and compressed about 23 to 40[\%] layered MQ information more than previous method. In future work, improve prediction accuracy that takes into account local area effect and the combination of MQ and use other entropy coding.
View full abstract
-
Masafumi ITO, Daiki SAKATA, Koji TASHIRO, Masayuki KUROSAKI, Hiroshi O ...
Session ID: 13-04-19
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we propose the joint source-channel coding (JSCC) algorithm for wireless image trans-mission using IEEE802.11ac system. JSCC is a technique that enables transmission of multi-layer and scalable data under bad channel condition. The proposed JSCC algorithm is a low-complexity subcarrier allocation on the11ac PHY layer. It improves image quality for a maximum of 20dB PSNR improvement compared to conventional method. Further, by constructing a demonstration system for simulating the image transmission in a wireless environment and applying the JSCC algorithm, show the utility of the proposed method.
View full abstract
-
- NON-ORTHOGONAL CODING USING THE SPACE-TIME CORRELATION -
Cui WANG, Yoshinori HATORI
Session ID: 13-04-20
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we proposed a new hybrid B-frame coding method which combined motion-compensated (MC) and non-causal prediction. The motion-compensated prediction is making full use of time correlation, while non-causal prediction is using the space correlation and has the characteristic of non-orthogonal transform coding. The new method is possible to reduce more MC prediction error signal power. Furthermore, in order to realize spatial replenishment control according to non-significant/significant pixel position of the input signal, a new conditional replenishment pixel control processing has been developed. This technology contains two parts: replenishment pixel decision in two kinds of resolution and differential feedback quantization to suppress the error in non-replenishment pixel region. According to these methods, fine grained rate control with maintaining the high coding performance achieved. Simulation results show that new coding method achieves about 1.0~2dB improvement in average when the entropy is almost the same to H.264 coding scheme.
View full abstract
-
Masayuki NAITO, Yoshinori HATORI
Session ID: 13-04-21
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
-
Hiromitsu KOBAYASHI, Kota AOKI, Hiroshi NAGAHASHI
Session ID: 13-04-22
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes a new method for extracting abnormal scene in the video of periodic machinery operations. Firstly, we segment the video into periods and extract STIPs (Space-time Interest Points). Then, we examine frequency of appearance of STIPs to determine regions of interest (ROIs). After the calculation of HOF (Histogram of Optical Flow) feature in the ROIs, we estimate abnormality of each segment using 1 class SVM (1 class Support Vector Machine). Experimental results show that the proposed method is sensitive to the variation of the machinery motion and robust against background and illumination variations.
View full abstract
-
Miki MIZUSHIMA, Yoshiaki TANIGUCHI, Go HASEGAWA, Hirotaka NAKANO, Mori ...
Session ID: 13-04-23
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Recently, pedestrian density estimation in crowded scenes based on optical flow extraction has attracted much attention. In this report, we propose a counting method of pedestrians passing through a specific line, called counting line, in crowded scenes. In our proposed method, optical flows whose segment crosses the counting line are used for estimation. In addition, we delete optical flows detected falsely or extracted from background image. Through experimental evaluations using actual video sequences, we show that the correlation coefficient between the number of optical flows and the number of pedestrians passing through the counting line is more than 0.856. We also show that the mean relative error of our proposed method is less than 0.068.
View full abstract
-
Takashi ETO, Yoshiki MIZUKAMI, Katsumi TADAMURA
Session ID: 13-04-24
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Efficient modeling of surrounding environments is necessary in provision of augmented reality services and productions of computer graphics works. This study discusses 3D shape reconstruction using a robot equipped with a depth sensor to reduce the time and effort required for the modeling. The robot has a stepping motor for changing the angle of the depth sensor and two DC motors for changing the robot’s location. It is necessary to estimate two parameters including the position and the angle in capturing each depth image to align the obtained depth images into a virtual 3D scene. We investigated two different methods to estimate the parameters. The first method calculates the parameters of the sensor from the amount of rotation of the actuators equipped on the robot. On the other hand, the second method calculates the parameters by applying ICP (Iterative Closest Point) to the depth images. These two methods were compared on the basis of experimental results.
View full abstract
-
Mitsushige KOMIYA, Hiroki TAKAHASHI
Session ID: 13-04-25
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes an intuitive navigation information presentation method based on analysis of indoor scene in order to ensure proper navigation information representation.3D point cloud data in a real space are measured by a depth sensor. A set of planes is detected from the 3D points. These planes are evaluated based on properties for navigation information visualization.Moreover, a color of navigation information is decided to facilitate intuitive understanding based on visual perception.
View full abstract
-
Yoshitaka TAKEMOTO, Kota AOKI, Hiroshi NAGAHASHI
Session ID: 13-04-26
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes a new spatio-temporal feature description method for human motion analysis of videos. First, cooccurrences of SOEs(Spatio-temporal Orientation Energy) between two different regions of whole body area are defined as a low-level feature vector. Second, some visual words are selected from these feature vectors obtained from various human motion video, and defined as templates. Then L2 norm between templates and a low-level feature vecter is defined as a middle-level feature vector. Experimental results show our proposed features can recognize human motions well by using one-all SVM.
View full abstract
-
Shun KITAHARA, Osamu UCHIDA
Session ID: 13-04-27
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In recent years, in sports such as football, baseball, volleyball, and tennis, quantifying various data including the movement of the ball and players is utilized to improve the players' skills and tactics of the match. However, in many cases, aggregating these data is still done manually. Therefore, a lot of time and effort is needed to record details of information. Equipment that can record various types of sports data automatically exists; however, the equipment is expensive in general and not everyone can use them with ease. In the previous study, we proposed a system to extract the players’ coordinate, the bound points of the ball, and the stroke points from tennis match videos automatically. However, it is required that both the start and the end time of each point (the sub division of the match between the time when one of the players serves the ball and the time when one of the players gets a point) are inputted into the system manually. Then in this paper we propose a method to recognize the scene of serving the ball automatically in order that the start time of each point will be inputted into the system automatically. Moreover, we propose a user interface that provides the player with useful information by visualizing various data extracted from recorded data comprehensively.
View full abstract
-
Yuuki Okada, Katsumi Konishi, Katsuyuki Shinohara
Session ID: 13-04-28
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This work focuses on a human identification by characteristic of the joint of body from defocuses and proposes a method. We use principal component analysis in measuring result from images to identify. Experimental results show that the proposed method can authenticate efficiently.
View full abstract
-
- Estimation From Hitting-Movements -
Piyarat Silapasuphakornwong, Masahiro Suzuki, Hiroshi Unno, Kazutake U ...
Session ID: 13-04-29
Published: 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper examines whether a new technique that we proposed to estimate the impact positions on the basis of human hitting-movements is applicable to a case where observers hit virtual objects presented with 3-D displays. In the proposed technique, which was examined under conditions where observers hit real objects, the impact positions were estimated based on the peak velocity of hitting-movements. In this paper, the velocity of hitting-movements was measured under conditions where observers hit 3-D images, and the proposed technique was examined using the data of this measurement. The results indicated that the proposed technique was able to estimate the peak velocity, and this suggests that the proposed technique is applicable to a case where observers hit 3-D images.
View full abstract
-
Takahiro FUKAYA, Hiroki TAKAHASHI
Session ID: 13-04-30
Published: February 20, 2014
Released on J-STAGE: March 31, 2019
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Gaze input enables us to operate devices without manual procedure.It is, however, hard to distinguish the two cases those are intentionally and unintentionally gaze. This is called Midas Touch Problem.An eye gaze based selection method, which employs saccade and screen buttons, is proposed to dissolve the Midas Touch Problem.The proposed method has wrong unintentional selection rate down to less than 1%.Moreover, dynamic calibration falls wrong selection rate from 21.7% to 5.0% caused by the measurement error.
View full abstract
