The Journal of the Institute of Image Electronics Engineers of Japan
Online ISSN : 1348-0316
Print ISSN : 0285-9831
ISSN-L : 0285-9831
Current issue
Showing 1-25 articles out of 25 articles from the selected issue
Foreword
In Memoriam
Report
Special Issue on Visual Computing
Contributed Papers
  • Makoto FUJISAWA (Member), Yuhei KATO, Masahiko MIKAWA
    2018 Volume 47 Issue 4 Pages 363-371
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    This paper propose a coastal terrain generation method based on a particle method and mathematical geography. The terrain generation is one of the important topic for computer graphics field and we focus on the coastal landform generation. In order to simulate the generation of coast, we have to consider a ocean wave. Most of previous terrain generation methods adopted simple water flow simulations to calculate the erosion, while it is difficult to accurately simulate the ocean wave because of its scale and its complex flow. Our method uses a low dimensional particle method to calculate the movement of the ocean wave and mathematical geography to simulate the change of terrain due to the wave. As a result, we archived a fast terrain generation for rock and sand coast.

    Download PDF (15377K)
  • Junya OKAWA, Yohei NAKADA (Member)
    2018 Volume 47 Issue 4 Pages 372-381
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    This paper proposes a 3D visualization tool for player and ball tracking data, which can be used for understanding of tactics in basketball. The proposed tool can reproduce a game by visualizing player and ball tracking data in 3D. In the proposed tool, the reproduced games can be observed from not only a bird’s–eye view but also each player’s–eye view. Thus, users can understand tactics effectively from a viewpoint of players in the games. The proposed tool can also display predictive information of pass targets based on players’ adjacent information. This predictive information can be helpful understanding tactics greatly, since pass plays are important in basketball. For a validation, the proposed tool is applied to player and ball tracking data included in an open dataset: APIDIS basketball dataset. Moreover, usefulness of the proposed tool is validated by based on the result.

    Download PDF (3012K)
  • Masataka EZOE, Hirohito SATO, Hironori MITAKE (Member), Shoichi HASEGA ...
    2018 Volume 47 Issue 4 Pages 382-390
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    Drawing dolls are used while drawing a illusration as a reference. It becomes good reference about proportion of human body. However, it does not help making natural human posture, which requires aesthetic sense of the creator. Human pose consists of several physical constraints such as forces to take pose and limitation of load in muscles or joints. Beginners of drawing human illustration often do not have this knowledge enough and make wrong posture which looks stressful. Therefore, we propose the virtual drawing doll with physics calculation, to suggest more natural posture based on users’ input of drawing doll’s posture. The system uses inverse statics to calculate joint torques, and calculates local optimal solution which requires less joint torques and maintains original posture as far as possible.

    Download PDF (14260K)
  • Naoki KATO, Masamoto TANABIKI, Junko FURUYAMA, Yuji SATO, Yoshimitsu A ...
    2018 Volume 47 Issue 4 Pages 391-400
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    One of the major factors that makes the task of tracking multiple objects difficult is frequent occlusions in congestion situations. Many existing methods in multi-object tracking by online processing have adopted a tracking-by-detection approach to perform object detection in every frame of a video and associate the object bounding box obtained by it temporally. However, with the existing method, it was not possible to track objects that could not be acquired by detectors due to occlusion. Therefore, we propose a method to shift the state of the lost target to tracked state once again by tracklet re-identification. Embeddings expressing the high dimensional appearance features of a object are acquired by using a convolutional neural network and re-identification of the tracklet is made based on the distance of the embedding vector among the tracklets. At this time, by using the masked image of the object obtained by instance segmentation as the input of the network, it is possible to perform re-identification determination robust to the background change. Further, since the re-identification determination of the tracklet pair is performed based on the distance between the low-dimensional vectors, the increase in the calculation cost due to the re-identification processing is small. Experiments were carried out using MOT16 dataset, which is a public data set, and the effectiveness of this method is shown.

    Download PDF (2592K)
Short Papers
  • Shohei YAMADA, Naiwala P. CHANDRASIRI (Member)
    2018 Volume 47 Issue 4 Pages 401-404
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    In many existing systems, the helper can show the annotations to the worker with drawn lines, but their display is planar. Therefore, it is unsuitable for expressing the depth information or instructing complicated work to the worker. In this research, we have devised a system that can deal with conveying of complicated tasks which involve depth information to the worker by using the hand gestures of the helper. In order to evaluate the usefulness of the proposed system, we conducted comparative experiments on remote work support by instruction annotations using conventional method of drawn lines and the proposed method of using hand gesture instructions respectively. As a result, no significant difference was found between two methods in terms of the understandability of the instructions, however regarding working time, the hand gesture instructions were shorter by 17% on average than the other method.

    Download PDF (1671K)
  • Akira SUEYOSHI, Ikuko SHIMIZU (Member)
    2018 Volume 47 Issue 4 Pages 405-409
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    Calibration method of cameras distributed in a wide area is important for many applications such as surveillance. Camera calibration method using motion barcodes is one of the methods for this purpose. Motion barcode is a time-seriese feature of a line in each image which describes the exisitence of the moving objects on the line. Methods using motion barcodes uses the correspondence of lines based on the simirarity of motion barcodes and correspondence errors are caused by the different moving objects existing simultaneously. In the proposed method, we introduce the color information to reduce the correspondence errors. In concrete, the correspondence errors are eliminated by comparing the hue histgram of the neighbouring regions of the moving objects whose colors are obviously different. In the experiments, our method and the conventional method was compared using the PETS2009 dataset and EPFL dataset.

    Download PDF (3575K)
  • Masaki ISHII (Member), Shinya FUJINO, Shunya SAITO, Rikuya TOGASHI, Hi ...
    2018 Volume 47 Issue 4 Pages 410-416
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    In this paper, we propose a leaving bed detection method using an RGB–D camera to prevent fall accidents when a care recipient leaves the bed. We classified the posture when leaving the bed into four types (supine position, sitting position on the bed, sitting position on the edge of a bed, standing), and attempted to determine the posture by using the height information and the moving distance of the head as features. In order to develop a method independent of the position of the camera, coordinate transformation was performed based on the floor of the room. As a result of the basic experiments, it became clear that the proposed method can distinguish four kinds of postures. The proposed method enables prediction and detection of behaviors when a care recipient leaves the bed.

    Download PDF (3425K)
Special Issue on VC2018
Contributed Papers
  • Tomohiro MIZOKUCHI, Yusuke TOKUYOSHI (Member)
    2018 Volume 47 Issue 4 Pages 418-424
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    This paper proposes a pipeline to accelerate the computation of frustum traced hard shadows. Recently this shadow algorithm has been applied to real-time applications such as video games, but it is computationally expensive compared to shadow mapping. To reduce the computation cost of the frustum tracing, this paper employs a two-pass visibility test by integrating a conservative shadow map into the pipeline of frustum traced shadows. Furthermore, this paper also presents a more precise implementation of the conservative shadow map than the previous method. In our experiments for 4K screen resolution, although the performance improvement varies depending on the scene, the shadow computation time is improved by about 2.4 times on average.

    Download PDF (2885K)
  • Naoki KITA (Student Member), Grégoire CLIQUET, Kazunori MIYATA (Member ...
    2018 Volume 47 Issue 4 Pages 425-432
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    We propose a mapping technique for projecting 2D plots onto spherical surfaces so that a user can observe and analyze data in an immersive environment. We project 2D plots that are as large as possible while minimizing distortions which covers a wider range of the user’s field of view. The data position surrounding the user which avoid an occlusion problem came from a visualization exploiting a conventional three-dimensional Euclidean space. The proposed method enables spherical mappings of high-dimensional data by first projecting them to a 2D space and then mapping to spherical surfaces and mappings of arbitrary 2D plots. We demonstrate application including high-dimensional immersive VR data visualization and decaling of spherical surfaces.

    Download PDF (4872K)
  • Tomofumi NARITA, Takashi IJIRI (Member)
    2018 Volume 47 Issue 4 Pages 433-439
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    This paper presents a semantic segmentation technique for three dimensional (3D) X-ray Computed Tomography (CT) images of natural objects, such as insects or plants. Our technique is based on knowledge that joints of semantically different parts of natural objects are often narrow. Given a binarized 3D CT image, we recursively detect the narrowest cross section that divides the foreground region into two parts. Our narrowest cross section detection consists of three steps; (i) splitting the foreground by erosion operations, (ii) regrowing the split regions by dilation operation and (iii) finding the narrowest cross section in the dilated region by adopting a graph cut method. To evaluate the accuracy of our technique, we adopt it to artificially generated images and found that error pixel rate was less than 2 %. To illustrate the feasibility of our technique, we adopt it to 3D CT images of insects and plants. As results, our technique successfully segmented multiple florets from an inflorescence, stems of a succulent plant, and legs of insects.

    Download PDF (5162K)
  • Shogo KISA (Student Member), Shigeru KURIYAMA, Tomohiko MUKAI
    2018 Volume 47 Issue 4 Pages 440-446
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    This paper proposed a motion synthesis and edit for character animations via latent variable space. An existing method using an auto-encoder is unsuited for efficient explorations and manipulations because the latent space has a large number of dimensions. Moreover, its encoder and decoder are composed of a single-layer, which is not suited to synthesize various types or styles of motions. We propose two types of generative neural networks that can map the latent variables so as to fit to a Gaussian distribution and can embed various motions in a lower dimensional latent space. We evaluate the plausibility of motions synthesized with our method, by demonstrating motion transitions and interpolations without preprocessing of time-alignment.

    Download PDF (1713K)
  • Yuji DEMURA, Makoto FUJISAWA (Member), Masahiko MIKAWA
    2018 Volume 47 Issue 4 Pages 447-453
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    In this paper, we propose a simulation method of human hair which enables plastic deformation such as a permanent wave and bed hair. Although simulation of hair is indispensable for expressing characters such as a human in the field of computer graphics, most previous simulation methods just treat the hair as an elastic body that is easy to simulate. Plastic deformation such as the influence of bed hair and hairdressing is not considered. In the proposed method, we introduce internal chemical bonds, in which actual hair plastically deforms based on its bonds. We simulate these bonds by repeating cutting and recombining process. By combining these with the position based method, a high-speed stable simulation has been realized.

    Download PDF (2617K)
Coffee Break
Contributed Paper
  • Lin MENG, Tsubasa TSUJI, Tomonori IZUMI (Member), Atsushi OCHIAI, Kats ...
    2018 Volume 47 Issue 4 Pages 457-466
    Published: 2018
    Released: November 30, 2020
    JOURNALS RESTRICTED ACCESS

    Oracle bone inscriptions (OBIs) are one of the oldest hieroglyphics which were inscribed on the bone of cattle or turtle shells in the period of Shang about 3000 years ago. Recognizing the OBIs is important for understanding the origin of characters, history research, etc. However, noise of OBI images, heterogeneity in inclination and size of OBIs made the OBI recognition difficult using image processing. OBIs are inscribed by sharp objects and consist of straight lines. Hence, we extracted the line features by Hough transform, and recognized the OBIs by calculating minimum distance between original images and template images. We proposed dependency matrix method for extracting the principal lines and calculate the minimum distance in the Hough space which is generated by hough trasform from the OBIs rubbing for recognizing the OBIs. We performed experiments utilizing 25 kinds, 550 pieces of OBIs and 35 kinds of templates. The experimental results show that 63 % OBIs were recognized in the first places, 78% in second place, 86% in third place.

    Download PDF (3702K)
International Conference
Technical Survey
Serial Technical Survey
Research Group Introduction
Scanning
Special Lecture
feedback
Top