ITE Technical Report
Online ISSN : 2424-1970
Print ISSN : 1342-6893
ISSN-L : 1342-6893
36.21
Displaying 1-12 of 12 articles from this issue
  • Article type: Cover
    Pages Cover1-
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    Download PDF (13K)
  • Article type: Index
    Pages Toc1-
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    Download PDF (36K)
  • Keiichiro KAGAWA, Shoji Kawahito, Jun TANIDA
    Article type: Article
    Session ID: IST2012-29/ME2012-80
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    TOMBO (Thin Observation Module by Bound Optics) is a compact three-dimensional camera, which is composed of a single image sensor and a micro-lens array. By utilizing a multi-channel image acquisition feature, expansion of the field-of-view is implemented with prism mirrors. Compact cameras with a wide, variable, or mixed field of view are mentioned.
    Download PDF (695K)
  • Yuzo Taketomi, Masayuki Ohata, Hiroshi Ikeoka, Takayuki Hamamoto
    Article type: Article
    Session ID: IST2012-30/ME2012-81
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    In this paper, we describe the improvement of horizontal focal planes method with the tilt optical system which is our conventional depth estimation method with a monocular camera. The conventional method estimates depth by taking advantage of the change of depth of field according to the depth using two or more horizontal and parallel focal planes. However, when estimating the depth of a far distant object, the interval of these focal planes becomes narrow on the image sensor plane under the influence of angle of field, and depth estimation accuracy substantially becomes worse. Therefore, in order to improve this problem, we propose making focal planes non-horizontal and non-parallel. It is possible to ease the restrictions on the interval narrowing of focal planes under the influence of angle of field, and it can be expected to improve the accuracy of depth estimation in the far distance.
    Download PDF (714K)
  • Tomoko MURAISO, Kayo OGAWA, Takashi KOMURO
    Article type: Article
    Session ID: IST2012-31/ME2012-82
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    In this research we propose a new motion classification method using principal component analysis to improve operability of a 3D gesture interface that assists text input on mobile devices. By applying principal component analysis to time-series finger scale data, the amount of information is reduced and only necessary information for classification is obtained. The result of experiment using a numerical keypad application showed that up to third principal components had eigenvalues more than one, and that the cumulative contribution ratio was more than 99%. The result of discriminant analysis showed that the proposed method achieved 94.3% recognition accuracy for mobile phone number input and 94.8% for circular constant input.
    Download PDF (846K)
  • Masana KIKITSU, Mitsuyasu ITO, Takayuki HAMAMOTO
    Article type: Article
    Session ID: IST2012-32/ME2012-83
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    In order to detect and avoid the obstacle, it is necessary for us to get the distance information in real-time. We have been investigating the distance estimation method using binocular camera with different focal position for each image sensor. This method estimates more precise distance by combining the information of multiple blur due to different focal positions and parallax information by binocular camera. In this paper, we implement the method on the camera system and evaluate the estimation accuracy from the perspective of the changes in the object size and the blur variation between images.
    Download PDF (641K)
  • Masahiro IWAHASHI, Hitoshi KIYA
    Article type: Article
    Session ID: IST2012-33/ME2012-84
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    A lossless layered coding for high dynamic range (HDR) images expressed in floating point format is proposed. It outputs two kinds of bit streams-a basic layer and an enhance layer. A low dynamic range (LDR) image represented in fixed point format with a standard bit depth is decoded from the basic layer bit stream. The original HDR image is decoded without any loss from both of the layers. Utilizing the lossless logarithmic tone mapping, the proposed method can reduces the bit depth of signal values in the enhance layer.
    Download PDF (661K)
  • Hiroyuki HASE
    Article type: Article
    Session ID: IST2012-34/ME2012-85
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    Human visual sense has two aspects in our feeling for blurred image, that is, one is the amount of blur depending on object size, the other is the amount of blur independent of the object size. In the former for example, when the image size becomes larger, we feel smaller amount blur. The quantitative evaluation based on entropy for blurred images is proposed in this paper. The author calls this metric "variation entropy". This metric has two kinds of aspects that coincide with the human visual sense. The first is the absolute evaluation of blur, and the second is the relative evaluation of blur. The former can be quantified by "variation entropy for a unit boundary length (or L-type variation entropy: H^L)", which is dependent on resolution, and the latter can be quantified by "variation entropy for a unit area (or A-type variation entropy: H^A)", which is independent of resolution. These two metrics have complementary properties. Last, two variation entropies are applied to the standard kanji character database, and then the strong relation between variation entropy and accuracy of recognition is discussed.
    Download PDF (613K)
  • Nathan McLean, Ken Tsutsuguchi, Yukinobu Taniguchi
    Article type: Article
    Session ID: IST2012-35/ME2012-86
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    We propose a technique to generate a vectorized representation of video data from raster video. Multiple frames from the input video are vectorized and used as keyframes and vector-based time data is automatically generated between pairs of keyframes. First, each keyframe is converted into a mesh constructed of triangular Bezier patches. Mesh vertices' are then tracked and approximated using Bezier spline curves. Vertices that were not successfully tracked were interpolated using a linear combination of nearby, successfully tracked vertices. We present the results of applying our method and discuss the validity and possible applications of the method.
    Download PDF (614K)
  • Article type: Appendix
    Pages App1-
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    Download PDF (86K)
  • Article type: Appendix
    Pages App2-
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    Download PDF (86K)
  • Article type: Appendix
    Pages App3-
    Published: June 04, 2012
    Released on J-STAGE: September 21, 2017
    CONFERENCE PROCEEDINGS FREE ACCESS
    Download PDF (86K)
feedback
Top