Reports of the 255th Technical Conference of the Institute of Image Electronics Engineers of Japan
-
Masateru MATSUO, Koji INOUE, Masayuki KUROSAKI, Yoshimitsu KUROKI, Bai ...
Session ID: 10-05-0
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this report, we propose a JPEG2000 parallel processing for 4K digital cinema wireless transmission system. Conventional parallel processing method using bit modeling order can not achive up to 30 frame persec.(fps). Proposed method can process in parallel using error resilience tools of JPEG 2000. Error resilience tools localize the error and make it tolerable for human eyes. We implement our proposed JPEG 2000 parallel processing by using compute unifed device architecture (CUDA) produced by NVIDIA. Then, we compare our proposed JPEG 2000 parallel processing and other reference of JPEG 2000 parallel implementation.
View full abstract
-
Yuya HIRATA, Masayuki KUROSAKI, Akio MIYAZAKI, Hiroshi OCHI
Session ID: 10-05-02
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this report, we propose an encryption and authentication system using Paillier encryption for digital cinema. We suppose the system is used in a home theater and the file format of movie is PEG 2000. The proposed system can check the ID in encrypted domain using Paillier encryption which has homomorphic mapping. In addition, the framework of proposed system conform to JPEG 2000 part 8(JPSEC) and therefore the proposal system has backward compatibility. Computer simulation explains the secrecy and the authentication of the proposed system.
View full abstract
-
Yuki HASE, Hiroshi ITO
Session ID: 10-05-03
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we propose a new watermarking method for intra-frames of H.264/AVC coded video. Intraprediction coding makes distortion by watermarking spread into next and following blocks. The spread produces prediction error variance increase and image degradation of watermarked digital contents. In the proposed method, coding reference blocks are given 1-point of weight when the block is used. Then, a watermark is preferentially embedded in 0-point blocks. By experiments, we show that this method prevents the prediction error variance increase and image degradation by the spread from being produced.
View full abstract
-
- Introduction of Our Researches -
Sakuichi OHTSUKA, Ken KIHARA
Session ID: 10-05-04
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This report introduces a part of our researches on the relationship between visual perception and information displays; (1) relationship between partial impairment and total quality, (2) relationship between composition and image quality induced by the change of aspect ratio in analog-digital simulcasting, (3) visibility of scrolled characters affected by human-body swing with direction dependency, and (4) “90-degree-hue-rotation method” which is a novel technique to help easy color discrimination for dichromatic vision preserving color saturation.
View full abstract
-
Kenta ASAKAWA, Tomonobu YOSHINO, Akira KUBOTA, Yoshinori HATORI
Session ID: 10-05-05
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we present a method of the super resolution reconstruction of B picture in a video coding. B picture in a video coding is low resolution as compared with I and P picture’s resolution. Therefore, as the resolution of B picture increases, so the resolution of the image sequences increases. So we proposed method of higher resolution for B picture. This is extended method of Iterative Back Projection (IBP method) for 3D. And we made it possible to high resolution reconstruction of one frame in image sequences.
View full abstract
-
Tomoya KURIMOTO, Akira KUBOTA, Yoshinori HATORI
Session ID: 10-05-06
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Recently, it is expected to find a more efficient encoding technology. In the H.264 encoding method, it has been used B picture encoding method by the casual interpolative prediction for high efficiency. And it is expanded to B picture encoding method to non-casual motion compensate prediction that is simple interframe prediction. However, for making predictive errors smaller and enhancing coding efficiency, we need introduction of a motion compensate prediction. But in this case, it has shown that a simple motion compensate prediction can not help to obtain efficiency for coding. In this paper, we proposed the motion compensate prediction to improve coding efficiency for non-casual interpolative prediction and the non-casual B-picture motion compensate prediction to cut the time of encoding and decoding.
View full abstract
-
Daisuke SUGIMOTO, Kenji MATSUO, Yoshinori HATORI
Session ID: 10-05-07
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We propose a new personal authenticaion system using video which is taken during sweeping handheld camera in front of one’s face. First, we defined Multi-dimensional subspaces from intrinsic images which had been derived from a registry video and an input video by subspace method. And we obtained a degree of similarity between two subspaces by mutual subspace methoc. Next, we translated a video data into a trajectory on the subspace for describing time-series information of the video. And we obtained a degree of similarity between two trajectories, one from registry video and the other from input video,by DP-matching. Finally, we authenticated two data from using two degrees of similarity,and achived EER=1.2%.
View full abstract
-
Hidekazu SEKIZAWA, Masahiro SUZUKI, Kazutake UEHIRA
Session ID: 10-05-08
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The miniaturization and making the color image reader an ultra-high-density are advanced. However, it is rigid conditions for the noise characteristic under the low illumination environment. It previously proposed the image input process when making the color image low noised had been attempted by confining it in a specific color space getting a hint from the vision model. This method is an excellent method to decrease the color difference noise with the characteristic of the ultra-high-density preserved. However, it was powerless for these ministers of RGB noise. On the other hand, the visibility rises in many of color images' different hue in the each composition part, preserving the edge part, and smoothing the inside.We detected the difference of the color by using the parameter when confining it in a specific color space, and developed the method to detect the edge part from this information. It has been understood to have a character excellent as the detection of the edge part by the halftone dot image of the printed matter etc. assumed that this proposal method is difficult especially so far becomes possible.
View full abstract
-
Kunio OHNO, Ken OU
Session ID: 10-05-09
Published: 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper describes a prototype system of the PIM (Personal Information Management) which has been standardized as iCalendar and / or vCard. Programing language LISP manages variable names through its symbol database. By means of this, PIM data of daily schedule, acquaint people, related document, etc., can be registered as special variables by setf macro and freely retrieved on its lisp listener. The schedule data can be converted to XML and displayed with daily, weekly, and monthly format. Finally, the possibility of PIM system on Web, based on S-equation has been proven.
View full abstract
-
Ken TSUTSUGUCHI, Mark DUPPENTHALER, Akira KOJIMA
Session ID: 10-05-10
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we describe our annotation editing system of vectorized images. Using a technique based on Xia et al., bitmap images are converted to vector images and stored as XML files. The user can load, display, edit the vectorized images and can add annotations to them. Added annotations are stored into the loaded XML file as XML tags. Our system can be applied to object-based image editing or rich contents creation.
View full abstract
-
Emi MYODO, Koichi TAKAGI, Shigeyuki SAKAZAWA
Session ID: 10-05-11
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Comic like layout, where multiple images are arranged in a screen while maintaining the image order like comic, has been proposed. For an effective presentation, panels on the screen should have variety in shape so that the resultant layout avoids tediousness. Therefore, conventional methods prepare templates which consist of horizontal and vertical divided lines of panels, cropped images in order to adapt the arbitrary rectangular-shaped panels in the templates. However, important areas such as regions of interests (ROIs) on images are often damaged in the cropping process under the constraint of maintaining image order. In the constraint, a series of required panel shape for image set is different per image sets, therefore preparing various templates is necessary. In this paper, we propose an exhaustive template generation method without appearance duplication. We employ rooted tree structures to express the templates with modification of the enumerating method of rooted ordered trees to keep the image layout balanced. The templates are evaluated based on shape fitness between ROI and panel. Simulation results show that almost all ROIs are included in the multiple layouts having high evaluation scores. Therefore, users can select a favorite layout among them. Moreover, the processing cost is low when the parameters are suitable for practical use despite of a large number of templates.
View full abstract
-
Hidehiko OKUBO, Mahito FUJII
Session ID: 10-05-12
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper presents a new method of the image completion of target region with information of local structures on adjacent regions of target. This method is one of our new technologies that utilize the additional information acquired on shooting, aiming to create more convenient and effective video composition and effect. We propose the completion method utilizes depth information that is used to calculate valid dataset regions to complete, and our experiment showed the possibility of improvement of synthesis result quality.
View full abstract
-
Yuya TANAKA, Ikuka SASAKI, Humiaki YASUOKA, Minoru ITO
Session ID: 10-05-13
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes a new approach for simply and precisely mapping a real texture of a two-dimensional (2D) object surface in a camera image onto surfaces of a model object and a virtual object in mixed reality space by using the 2D cross-ratio technique. The object surfaces treated here are planar quadrilaterals with four vertices. The 2D surface texture mapping is based only on the cross-ratios among the vertices. The main advantage of the proposed approach is that the three-dimensional perspective projection transformation inevitably required in the conventional methods is not needed at all, which means the principle and procedure for the mapping are very simple. A real-image experiment demonstrates the effectiveness of our proposed approach.
View full abstract
-
Yusuke TAKEI, Kota AOKI, Hiroshi NAGAHASHI
Session ID: 10-05-14
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes a method of motion generation a non-rigid object based on the field of reaction-diffusion equations and principle of locality. The object surface composed of piecewise polynomial patches deforms it's shape autonomously and generates a natural motion as a whole. Complex motions can also be generated by controlling several parameters that define the potential field represented by the reaction-diffusion equations. The generating object motion is performed by studying the best control parameters that maximize a certain evaluation function.
View full abstract
-
Makiko NAGASAWA, Yoshinori HATORI, Mitsugu KAKUTA, Tadao HAYASHI, Yosh ...
Session ID: 10-05-15
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Motion Analysis for Badminton Smash based on Biomechanics Information is proposed.
View full abstract
-
Tatsuya OBATA, Kota AOKI, Satoshi FUJITA, Hiroshi NAGAHASHI
Session ID: 10-05-16
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In a research of microscopic image processing such as a stem cell study or drug development, it has been necessary to extract features and trace them manually, because the shape and size of cells greatly change. In this paper, we propose a system which can perform extraction and tracking of the cell precisely in complicated microscopic images. This system identifies a cell by two consecutive images by using features of the cell such as the coordinates of the center of gravity, area, shapes, intensity and velocity. We show that the proposed combinatorial optimization method by limited parts is effective for cell tracking.
View full abstract
-
Qian WANG, Kota AOKI, Hiroshi NAGAHASHI
Session ID: 10-05-17
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In actual surveillance conditions, there exist a lot of uncertainties in pedestrian movements. These movements may disturb most tracking algorithms and result in tracking failure. In this paper, a new pedestrian tracking system is proposed, in which an online-boosting method is embedded into the pedestrian tracking of object model. Online-boosting [1] methods have become very popular in computer vision, and achieved impressive performance in detection and recognition tasks. However, the on-line adaption has to face one key problem: an error which finally leads to tracking failure (drifting) may be accumulated in each update of the tracking. In order to avoid the drifting problem, MIL (Multiple Instance Learning) method is proposed, instead of traditional supervised learning, to lead to a more robust tracker with well performance.
View full abstract
-
Fumiyuki MATSUSHIMA, Kota AOKI, Hiroshi NAGAHASHI
Session ID: 10-05-18
Published: February 24, 2011
Released on J-STAGE: January 13, 2015
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes a technique of detecting wires on a printed wiring board(PWB) in a noisy image produced by x-ray CT. In the proposed method, we extract line features by using gabor filter and detect wires by these features. Firstly, Using gabor filters with various inclinations, we detect segments in different directions and produce binary images. Secondly, these produced images which contain a lot of noises are filtered in order to remove noises. Finally, we connect lines which become interrupted. Experimental results show that most wires can be detected.
View full abstract
