-
- Composition of the Effective Parts of Multiple Exposure Images -
Takaharu YAMAMOTO, Masayoshi AOKI
Session ID: 03-07-01
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Because of limited dynamic range of film and imaging device, wide dynamic range scene cause black out or white out. In this paper, we propose a scheme to improve dynamic range, by combining multiple exposure images of a wide dynamic range scene. This scheme, selects parts of multiple exposure image which contain effective image information, and combines them into an image free of black out and white out. Results of applying the proposed scheme gave natural looking image without any perceivable boundary between combined image parts.
View full abstract
-
Yuki OSAKI, Shin'ichi TAKAGI, Akihisa KODATE, Hideyoshi TOMINAGA
Session ID: 03-07-02
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We have been doing a study on a low-cost high resolution digital image acquisition system based on image mosaicing technique. Our system assumes that image frames are taken with a digital video camera, and that would cause color differences between image frames, and shading in a single image frame. Therefore, we propose to capture a color chart, which the original color is known, with the object, to correct the color differences in-between image frames. This paper is focused to describe the method to correct shading in an image frame. We would show the effectiveness of our method through experiments at the end.
View full abstract
-
Shoichi YOKOI, Tomohiro HORIUCHI, Tahseen EJAZ, Yoshifumi SHIMODAIRA
Session ID: 03-07-03
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we describe a camera that can realize the colorimetric color reproduction and get the image with XYZ tristimulus values. We want to build the camera that uses single photo acceptance unit and receives light from 3 different filters simultaneously. So, we discuss the ideal sensor response functions for it, and check the quantization error for SOCS reflectance database. We made a prototype that has 3 bands with non-negative sensor response functions. Experimental results show good accuracy of LAB delta E units for the Macbeth color chart under an artificial sunlight.
View full abstract
-
Emi TANAKA, Emi FUKASAKU, Miwako SUZUKI, Yasuo KUHARA, Noboru MURAYAMA
Session ID: 03-07-04
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We developed a proto type of the Aroma-Printer by replacing the ink cartridge of the inkjet printer with the aroma-cartridge. The aroma material is different from the ink not only in the chemically hydrophilic or hydrophobic aspect but also physical property of specific gravity, surface tension and so on. In the prototyping of aroma-printer, we experienced many problems and solved the problems step by steps, by try and error base experiment. We applied the prototype aroma-printer to the aromatherapy experiment. Some of the problems are not solved and still remain for the future. The biggest future problem may be the clogging problem as in the ordinary ink jet printer.
View full abstract
-
Tomonori KANEKO, Yuichi ASADA, Seiichirou TOKIOKA, Minoru ITO
Session ID: 03-07-05
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes a new approach for detecting rotations and scalings of images, where our previous phase difference spectrum analysis (PSA) method is applied. In the experiment, three typical image patterns are rotated and scale-changed synthetically or by a mechanical stage. For synthetically rotated images, the averaged absolute error is 0.006 degree, while it is 0.053 degree for images rotated by the stage. In detection of scale changes, the absolute error is 0.008 in average.
View full abstract
-
Kousuke HARA, Minoru AKUTSU, Tetsuya OOTANI, Takashi OMORI
Session ID: 03-07-06
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The production of flat panel displays and image sensors has been increasing rapidly by the recent diversification of their uses. For that reason, in the defect inspection of those image devices, the automation of image inspection has proceeded. However, it was difficult to make general procedures of image processing, because there were various defects. Therefore, engineering cost increases. This paper describes a searching system for image processing procedures and parameters. In particular, we propose the fast method using the searching results for similar defects.
View full abstract
-
Soichiro IKEO, Norihiko ITOH
Session ID: 03-07-07
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We have developed Analog to Digital converting system using image processing for telemetring of oil thermometers and pressure meters. Practical use of this system may realize very effective integration of analog meters to digital based control system.Now this system is evaluated as a viewpoint of application to Automatic surveillance systems in which as yet many traditional analog meters are used.
View full abstract
-
Tomoyuki FUJIHARA, Makoto OHKI, Sumihisa HASHIGUCHI
Session ID: 03-07-08
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In tracking a moving object by a panning camera three images were taken to detect moving object by the camera at rest. The differences of the taken images gave only the traces of moving object, without any traces of the background at rest. With the algorithm the target was successfully traced.
View full abstract
-
Atsushi ITO, Shuji HASHIMOTO
Session ID: 03-07-09
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes two methods to process 3D shape data. One is a filtering method for 3D shape to extract global and local information of 3D shape by the wavelet transform of range data from the origin of polar coordinates. Another is a statistical method for automatic registration that can integrate 3D shape data with different numbers. It is realized by the affine transformation to minimize the difference of moments between the data sets. Furthermore, more robust registration is performed by combining shape filtering method and statistical registration method.
View full abstract
-
Tsuyoshi TAKEI, Ryutarou OHBUCHI
Session ID: 03-07-10
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We have previously published a multiresolution approach for the shape similarity comparison of polygon soup 3D models. The approach combined a single-resolution shape feature with a multiresolution representation produced by using the 3D alpha shapes algorithm of Edelsbrunner for the shape similarity comparison. The study evaluated the efficacy of the multiresolution shape comparison approach by using only one shape feature and only one database. This paper attempts a more comprehensive evaluation of the efficacy of the multiresolution approach by using four different single-resolution shape features and two different databases. Experiments showed that (1) the multiresolution approach does improve the performance of all the four shape features tested, and (2) the combination of the Surflet Pair Relation Histogram with the multiresolution approach performed the best among the tested.
View full abstract
-
Yuushin HATA, Akira HORIUCHI, Tsuyoshi TAKEI, Karman CHAN, Ryutarou OH ...
Session ID: 03-07-11
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The popularity of 3D shape models in 3D games, 3D mechanical CAD, and other applications lead to explosion in the number of 3D models, and to recent interest in 3D shape similarity search and retrieval. Previously, Nakazawa, et al reported a shape similarity comparison method that computes shape similarity by using a set of 2D images of a 3D shape taken from multiple viewpoints. This paper follows up on their work, presenting an improved distance computation method and a detailed evaluation of the influence of the numbers and positions of viewpoints to the shape comparison performance.
View full abstract
-
Koichiro TSUJI, Akinori MURANAKA, Shigeru IMANISHI
Session ID: 03-07-12
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Multi-Valued logic system is introduced into the important preprocess of image recognition, and extracts image contour with it difficult to extract in processing of 2 value system generally used. Moreover, we compare our Multi-Valued logic circuit we designed with usual 2 value logic circuits. the miniaturization of circuits by curtailment of wiring count and transistor count ,and examine.
View full abstract
-
Gongyu SUN, Kui ZHANG, YueFeng LU, Masaki OSHIMA
Session ID: 03-07-13
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes automatic image recognition of hand-over action between two persons. In this research, the three kinds of action states, holding an object in a hand, separating the object from the hand, transferring the object into another hand, will be automatically identified. For this purpose, first, the persons are extracted from background using background difference, second, the hands of persons are detected by the skin-color, third, the object is identified based on the size of moving range related the hands, fourth, the motions of the hands and the object are tracked using block matching. Finally, using the change of relation between the hands and the object, the hand-over action is recognized. At all, the recognition technique of hand-over action was proposed, and it experimented as a computer program. In the laboratory environment, the validity and the effectiveness of this system is shown.
View full abstract
-
Shunya SUZUKI, Mei KODAMA, Tomoko TANAKA, Saori TSUNEKUNI
Session ID: 03-07-14
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In the VOD service, the low resolution video data is expected for content searching, and the hi resolution video data is expected for browsing. In our previous studies, we presented the effect of scalability encoding in cache utilization. In this study, we report comprehensive experiments on cases of various updating procedures, and the effects of differencial data transfering. As a result, if a time-cycle of downloading, transcoding and browsing is too short for all clients to download contents within, the scalability encoding method can improve the quality of service, in both of data size and consumed time.
View full abstract
-
Aya KATO, Eriko WATANABE, Kashiko KODATE
Session ID: 03-07-15
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The cellular-phone face recognition system is proposed as an alternative and a more reliable participant recognition system. It is designed and constructed, utilizing our original system; Compact Optical Parallel Correlator (COPaC), to accelerate the whole process of capturing, recognition and identifying images. The system contains photographs with a cellular-phone camera and biometrics recognition under a mobile environment. By proof experiment using 20 subjects, the rate of recognition 95% confirmed efficiency of the working system and its practical use.
View full abstract
-
Yasunobu OGURA, Toshiaki EJIMA
Session ID: 03-07-16
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We investigate the architecture of human face detector which makes"Coarse to Fine" classification processes work well. Measuring accuracies and computation costs for various types of classification architecture and comparing them, we clarify relationship between detection architecture and classification functioning. As a result, T architecture is superior to the others, which is able to reject most of non-face image quickly at the first or the second stage using a simple classifier and detect face correctly at final stage using several classifiers in parallel.
View full abstract
-
Nana TAKEDA, Toshiaki EJIMA
Session ID: 03-07-17
Published: March 12, 2004
Released on J-STAGE: April 07, 2004
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
We investigate human identification using motion pattern of leg (MPL). We utilize Fourier transform of MPL to identify individual among a set of human. Through experiment for human identification, it has been shown that amplitude spectrum of MPL is superior to phase spectrum as a feature for identification. In addition to that, using complex form as representation of two MPL, that is, MPL of right and left legs leads to better performance than using two simple MPL. As a result, using complex form of MPLs, we have achieved identification accuracy over 83.4% in telling who one is of four persons.
View full abstract
