Reports of the Technical Conference of the Institute of Image Electronics Engineers of Japan
Reports of the 233rd Technical Conference of the Institute of Image Electronics Engineers of Japan
Showing 1-14 articles out of 14 articles from the selected issue
Date: August 1, 2007 Location: Nagaoka University of Technology
9:25-10:40 Chair: Hiroshi YOSHIKAWA, Nihon University
  • Masaki HIRAOKA, Kousuke IMAMURA, Hideo HASHIMOTO
    Session ID: 07-01-01
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Extraction of semantic video objects is an indispensable tool that benefits newly developed video description schemes. In the present paper, we propose a technique that extracts the shape of moving object by combining snakes and watershed algorithm. The proposed method comprises two steps. In the first step, snakes extract contours of moving objects as a result of the minimization of an energy function. In the second step, the conditional watershed algorithm extracts contours from a topographical surface including a new function term.
    Download PDF (468K)
  • Sakol UDOMSIRI, Natanit BORIBOONHIRAN, Marielena Palacios PEREZ, Masah ...
    Session ID: 07-01-02
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Nowadays, various types of the “ integer ” DCT which outputs integer numbers have been widely investigated and its performance have been continuously improved for being able to be used in many areas such as data compression. This report evaluates performance of various integer DCTs and finds the best integer DCT structure from all the possible combinations of various rotation transforms. As a result, it is found that the best integer DCT structure gives the least sensitivity for the finite word length expression of multiplier coefficients and improves PSNR up to 4.29 [dB] for the AR(1) model as an input signal.
    Download PDF (465K)
  • Makoto OHKURA, Shogo MURAMATSU, Hisakazu KIKUCHI, Masahiro IWAHASHI
    Session ID: 07-01-03
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    This report proposes a river region classification using a Gaussian mixture model for feature vectors. The conventional method classifies river region by assuming a single Gaussian distribution. Since this work assumes a Gaussian mixture distribution, better classification performance can be expected because it becomes possible to approximate more complex distributions than any single Gaussian. Experimental results with real surveillance videos are shown for evaluating the performance of the proposed method, and improvement of the performance from the conventional method in term of precision is verified.
    Download PDF (636K)
10:50-12:05 Chair: Fumitaka ONO, Tokyo Polytechnic University
  • Hisashi OWADA, Madoka HASEGAWA, Shigeo KATO
    Session ID: 07-01-04
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    The Motion Compensation technique and Discrete Cosine Transform are used in the standardized moving image coding method MPEG. In the Motion Compensation, an image is divided into macroblocks, and the macroblosk size of 16x16 pixels is used in the MPEG2, and 16x16 to 4x4 pixels in H.264/MPEG4-AVC. However, when the Motion Compensation is applied to the high definition videos, the size that is larger than 16x16 pixels might become efficient. In this paper, we examined the suitable macroblock size in the Motion Compensation to the high definition videos.
    Download PDF (139K)
  • Masahiro TSUJI, Madoka HASEGAWA, Shigeo KATO
    Session ID: 07-01-05
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Recently, digital technology is advanced in many fields, and demands that the paper media such as the newspaper or the magazine are linked with the digital information will tend to increase in future. Various two-dimensional codes are proposed as techniques to link the digital information from the analog contents. However, two-dimensional code is meaningless information for human eyes, and it wastes space of paper. In this paper, we propose a method to solve this problem by applying digital watermarking to printing images. This method embeds a watermark in a monochrome image by using brightness value of pixel blocks.
    Download PDF (369K)
  • Shunya HATAKEYAMA, Madoka HASEGAWA, Shigeo KATO
    Session ID: 07-01-06
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Owing to development of the printing and scanning technology the opportunities to treat a mixed image of the photograph and character are on the increase. However, the file size of high-resolution document image is usually large, and costs of communication and storage increase. Therefore a document image is classified into a character region and photograph region, and encode them by the suitable method for each region. In this paper, we propose a classification method into a character region and a photograph region from the mixed image by using Markov model contexts.
    Download PDF (248K)
13:00-13:30 Chair: Masahiro IWAHASHI, Nagaoka University of Tevhnology
13:30-14:45 Chair: Hiroshi OCHI, Nihon University
  • Suguru SAIKI, Ken'ichi TANAKA
    Session ID: 07-01-08
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Computer-generated hologram is made for three dimensional image reconstruction of a virtual object. Error diffusion method is used to improve the reconstructed image. The selection of diffusion coefficient exerts an influence to the image quality of the reconstructed image. Therefore, many researchers study the optimization of coefficients. In this article, estimation of coefficients using statistical analysis is done. Using regression analysis, it's possible to relation between coefficients and object position. So, processing time is very short and reconstructed image quality is well. This paper presents the suggested method and the reconstructed image.
    Download PDF (264K)
  • Tomotaka NAKANISHI, Katsuyuki SHINOHARA
    Session ID: 07-01-09
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    At present when a huge amount of image can be maintained, the image retrieval that searches out the image of the target is very important. Then, the retrieval technique that uses the multiple resolution analysis by the Wavelet transformation is examined in the main discourse. As for one time, a huge computational complexity is necessary for the multiple resolution analysis. Then, other blocks are disregarded for the block with the feature. And, data having been made a dimension by scanning to the block with the feature zigzag one. The face image was enabled to be retrieved in the data base by converting one dimension data in Wavelet, and using the obtained each element.
    Download PDF (443K)
  • Takuya ABE, Katsuyuki SHINOHARA
    Session ID: 07-01-10
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In the recent research of similar image retrieval, the method computing the degree of similarity using the combination of local features which are extracted around the interest points attracts attention. However, it is hard to extract interest point from the scene images, and search results are affected by it. The authors are researching feature extraction by using wavelet filter. In this case, we set threshold manually. So we examined the method of setting the threshold, and evaluated this technique in this report.
    Download PDF (271K)
14:55-16:10 Chair: Hidekazu SEKIZAWA, Toshiba Tec Document Processing Systems Co.,Ltd.
  • Masaki TANABE, Katsuyuki SHINOHARA
    Session ID: 07-01-11
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    We propose the similar image retrieval using PCA(principal component analysis) in this paper. PCA is a kind of multivariate analysis technique and it can concisely summarize a lot of variables. In our technique, to apply the PCA to the similar image retrieval, PCA was done to the image which is smoothed and grid-divided based on the cumulative contribution of the eigenvalue in each grid. So we are aimed at extracting unique information from the image and searching for similar images using main information of the image.
    Download PDF (382K)
  • Mitsuo HAYASHI, Nobuhiko MUKAI, Makoto KOSUGI
    Session ID: 07-01-12
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    We have examined the difference of distances in virtual and real spaces by comparing the distance from the eye position to the virtual and real objects. As the result of the experiment, it was found that we tend to recognize an object further than the real distance when we see the object which is nearer to the eye position and vice versa. Next, we have measured the real distance from the eye position to a virtual object without comparing real and virtual objects. As a result, it was found that we tend to recognize objects further than the real distance and its feature is remarkable when we see objects which are nearer to the eye position.
    Download PDF (354K)
  • Ayako SUGAWARA, Ryo YAMAMOTO, Daisuke ENDO, Masanobu YAMAMOTO
    Session ID: 07-01-13
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    Location hunting, which is very important issue in film making, usually sped much cost for choosing the best location. This paper proposes a novel system that can hunt the best location in virtual environments. We take images of locations from multiple-view whose position and orientation are measured using GPS, an electronic compass and angle meters. The 3D environment of the location can be reconstructed from these images. Comparing movie shootings in the several virtual environments, the user (a director of film making) can choose the best location.
    Download PDF (508K)
16:20-17:00 Chair: Fumitaka ONO, Tokyo Polytechnic University
  • Masahiro IWAHASHI
    Session ID: 07-01-14
    Published: 2007
    Released: December 18, 2007
    CONFERENCE PROCEEDINGS RESTRICTED ACCESS
    In this report, a new type of layered (scalable) video coding especially for the use of water level monitoring is introduced. A sensor node of the system produces a bit stream functionally separated into three layers. The first layer contains the minimum components effective for detecting the water level. It is transmitted at very low bit rate for regular monitoring. The second layer contains signals for thumb-nail video browsing. The third layer contains additional data for decoding the original video signal. A video signal is decomposed into several bands with the three dimensional transform. In this report, optimum bands to be contained into the 1st layer are experimentally investigated considering both of water level detection and data size to be transmitted. As a result, bit rate for transmitting the first layer is reduced by 32.5 % at the cost of negligible 3.7 % decrease of recognition performance for one of video examples.
    Download PDF (682K)
feedback
Top