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
Displaying 1-18 of 18 articles from this issue
Foreword
Special Issue on Internet of Things and its Related Technologies in Image Electronics
Contributed Papers
  • Naoto KAWAMURA (Fellow)
    2019 Volume 48 Issue 2 Pages 221-227
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    We propose an invisible block type Digital Watermarking Method which enables to embed multiple information. The main watermark information is embedded by Green-Noise dot pattern, and secondary information is superimposed on it. The Green-Noise pattern is perceptually low visibility for human visual system at the distance of distinct vision, and it represents high resilience for various attacks. The secondary information is embedded with a higher frequency pattern which does not overlap with the spectrum of main information, then this will result in a fragile watermark. Embedded information can be extracted separately by different extractors. The possible applications are copyright protection, detection of tampering, confirmation of original, tracking of illegal outflow etc. Furthermore, the sub-information can be used as a copy-forgery-inhibited pattern and as a marker for block synchronization.

    Download PDF (8482K)
  • Kentaro YAGI, Kunihiro HASEGAWA, Hideo SAITO (Member)
    2019 Volume 48 Issue 2 Pages 228-236
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    In recent years, techniques for diminishing objects from images are actively researched. These technologies are used for multiple applications such as privacy protection for people who are unconsciously in images. However, conventional methods have constrains such as “multiple cameras required”, “the objects to be diminished is required to be static”, “the background must be a plane.” In this research, we propose a method for diminishing people in an image sequence using a single handheld camera. We suppose that the scene is dynamic of which both photographer and people to be diminished are moving while the video is being recorded. In additon, the background can be arbitrary shape because we reconstruct 3D shape of it. In the experiment, we compare the proposed method with existing methods for different scenes which have various backgrounds.

    Download PDF (28334K)
System Development Papers
  • Shinya HATSUDA, Masashi ONO, Lin MENG, Tomonori IZUMI (Member)
    2019 Volume 48 Issue 2 Pages 237-247
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    Aiming for anti-vermin surveillance cameras, this paper presents evaluation and trial-implementations of classifiers based on machine learning and construction of database to improve the accuracy for actual images. Based on a database composed of images taken in a zoo, we construct classifiers with HOG features and a support vector machine (HOG + SVM), HOG features and a neural network (HOG + NN), and a convolutional neural network (CNN). The results show that the HOG + SVM classifier is the fastest and the CNN classifier achieves the best accuracy. Trial implementations of the two classifiers on an embedded processor show that the HOG+NN achieves 88.5% accuracy and takes 8.7sec processing time for a VGA sized image frame while the CNN achieves 94.5% accuracy and takes 45.8sec processing time. For images actually taken in shrines and temples, an experiment reveals that the CNN achieves merely 20-30% of accuracy but that training with a database composed of differently characterized image sets improves the accuracy to around 60%. An improvement of the program with background subtraction technique is given and the processing time is reduced to few seconds or less. The results indicate that the CNN trained with the carefully constructed image database shows good accuracy to detect raccoons on low-computing-performance surveillance cameras.

    Download PDF (2042K)
  • Yoshihiro SUGAYA, Kiyoshiro SAKAI, Tomo MIYAZAKI, Shinichiro OMACHI (M ...
    2019 Volume 48 Issue 2 Pages 248-257
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    Smartphone Apps that recognize texts captured with a camera and provide information based on the recognized texts are becoming widespread. However, they are hard to be used in situations where hands cannot be freely used, and also for handicapped people. There are smart glasses as a device suitable for the information providing system under such situations, however, there were no items that had the performance necessary for character recognition in the scene image and did not interfere with work and actions. In this paper, we developed a wearable system for information providing applications using scene text recognition. The prototype device has a camera suitable for character recognition in the scene image and a display that does not hinder the tasks and actions while wearing. In addition, it has a function to communicate with an external PC or a cloud server, and APIs for that are also developed. Furthermore, since it has no input device except for the camera, we developed a user interface suitable for such a device. We implemented a sample application that recognizes a station name on the route map and displays information on the station. We conducted evaluation experiments of the user interface and the whole system, and confirmed that it is practical for tasks that are allowed to retry.

    Download PDF (3470K)
Short Papers
  • Kunihiko TAKANO (Member), Syuntaro TOMINAGA, Tatsuya IKEUCHI, Koki SAT ...
    2019 Volume 48 Issue 2 Pages 258-262
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    As an effective reconstructing technique to observe full-parallax holographic images in the wide viewing area, we have studied a projecting process of 3-D holographic images with the use of the spatial screen. As one technique to perform a suitable projection of the images, projection of holographic 3-D images onto the underwater micro-bubble screen is well known. Applying this technique, the stability of the images displayed in the screen seems to have been highly improved, however, as for the flickerling of the images, much more improvement seems to be not so easy. For this reason, we constructed a new spatial screen system employed fog filled in the tank. In this report, we have discussed its effects to the reconstructed images through the holograms. We have seen that effective for suppressing the flickering of the images with higher resolution.

    Download PDF (4655K)
Special Issue on Interaction Technologies between Image Electronics and Real World
Contributed Paper
  • Munekazu DATE, Daisuke Ochi, Hideaki KIMATA (Member)
    2019 Volume 48 Issue 2 Pages 264-272
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    A novel autostereoscopic flat-panel 3D-display, which can display viewpoint images directly, that can reproduce smooth horizontal motion parallax using small number of directional images is introduced. Using the human perception characteristics of doubled images, we interpolate directional images and generate visually-equivalent rays using optical linear blending in the display. Our previous proposal required a huge apparatus and had low image quality. Since this new display requires only that a barrier, whose spacing is almost same as pixel width, be added to a regular horizontal RGB stripe type 2D LCD, it yields thin flat panel displays. The lens-less optical configuration enables high image quality, because it is intrinsically free from distortion and blur. Moreover we sorted out phenomena from a standpoint of spacial frequency and showed that only assuming cutoff frequency of visual system, perception can be explained optically.

    Download PDF (1994K)
Short Papers
  • Yasutomo KAWANISHI (Member), Daisuke DEGUCHI, Ichiro IDE, Hiroshi MURA ...
    2019 Volume 48 Issue 2 Pages 273-277
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    In this paper, we propose the concept of “Person Re-detection”, which is a person detection method based on prior knowledge of each detection target. We also introduce an implementation of the concept for the task of distant person detection from in-vehicle camera video. We focus on the fact that some other vehicles around the own vehicle can observe and detect people before the own vehicle observes them. By sharing images of the detected people among the vehicles, the own vehicle can use them as prior knowledge for detecting them. Results of applying the proposed method to a dataset obtained by an in-vehicle camera demonstrate that the accuracy of person detection can be significantly improved based on

    prior knowledge of the people to be detected.

    Download PDF (1727K)
Coffee Break
Contributed Paper
  • Shu FUJITA, Sho MIKAWA, Keita TAKAHASHI (Member), Toshiaki FUJII (Memb ...
    2019 Volume 48 Issue 2 Pages 281-289
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    The multi-focused plenoptic camera is a powerful device that captures an light field (LF), which is interpreted as a set of dense multi-view images. The camera has potential ability such that we can obtain LFs having high spatial/view resolution and deep depth-of-field. Due to its effectiveness, this camera design attracted attention in vision and graphics fields. To extract multi-view images, we need a sophisticated rendering process due to its complicated optical system. However, there are few studies on this. Furthermore, the only available rendering software to the best of our knowledge has not worked well in several camera configurations. We therefore introduce an improved rendering method and will release the software that uses the method. Our software can produce better LFs than the previous one and robustly work in various camera configurations. We thus believe that our work can contribute to the computer vision and graphics community.

    Download PDF (8356K)
  • Naoto KAWAMURA (Fellow)
    2019 Volume 48 Issue 2 Pages 290-302
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    We propose robust digital watermarking technology based on green-noise diffusion method, which represents high resilience to printing. As the spectrum of green-noise pattern is confined between fmax and fmin, the MTF response is high in the printer, but low in the human visual system at the distance of distinct vision. For this reason, the watermark information is retained while keeping high image quality. Watermark embedding is performed using two different patterns, which corresponding to bit 0 or 1, by adjusting the embedding strength (gain) while looking at the reliability. Since the spectrum of green-noise pattern and the image are less overlapped, the watermark information can be extracted in the spectrum domain. Also, since the embedded pattern is diffused by random dots, it cannot be easily removed, but only removed with a dedicated “key”. Since this key can be issued indefinitely, even if it is stolen it cannot remove watermarks of other images. Therefore, this system is highly secure. Since confidentiality is aggregated into this key, it does not promote attacks from third parties even if the algorithm is disclosed. In addition, as a method of block synchronization which is a significant problem in block type, it is possible to synchronize with high accuracy by extracting “marker” in which a pattern not overlapping the Green-Noise spectrum is embedded. In this paper, two applications of copyright protection and annotation were investigated.

    Download PDF (5485K)
  • Takuya KATO, Satoru FUKAYAMA, Tomoyasu NAKANO, Masataka GOTO, Shigeo M ...
    2019 Volume 48 Issue 2 Pages 303-314
    Published: 2019
    Released on J-STAGE: February 10, 2022
    JOURNAL RESTRICTED ACCESS

    In three-dimensional computer-graphics speech animation, head rotations and facial features other than mouth motions are considered to be equally important as mouth motions. Such motions especially in singing animation is known to affect the impression of the viewers and requires a huge amount of work by hand when creating them. Therefore, there is high demands in creating such animation with available inputs. In this work, we present singing animation synthesis method with the information associated with the song. Using singing voice, musical information and the mouth motion estimated from the singing voice as input, the proposed method can automatically create realistic head motion and expressions around eyes. We have conducted the numerical evaluation as well as subjective evaluation on the naturalness and the synchronization to the singing voice and song to show the effectiveness of our method.

    Download PDF (4304K)
Serial Technical Survey
Research Group Introduction
Scanning
feedback
Top