ITE Technical Report
Online ISSN : 2433-0914
Print ISSN : 0386-4227
Volume 18, Issue 15
Displaying 1-13 of 13 articles from this issue
  • Article type: Cover
    1994 Volume 18 Issue 15 Pages Cover1-
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    Download PDF (12K)
  • Article type: Index
    1994 Volume 18 Issue 15 Pages Toc1-
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    Download PDF (53K)
  • Mamoru TSUCHIYA
    Article type: Article
    1994 Volume 18 Issue 15 Pages 1-8
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    A new design technique for FIR( Finite-duration Impulse Response ) digital filters with a small number of taps was developed. It is suitable for digital processing of TV picture signals and this algorithm is an analytical design method, with, which tap-gain coefficients of FIR filters are obtained by non-iterative calculation. In addition, a minimum stopband attenuation can be reduced within proper values by permitting equi-ripples on the passband without extending the transition width of FIR filters. A method to reduce the level of equi-ripples is also mentioned.
    Download PDF (612K)
  • Shin-ichi NAGAIWA, Eliana Eriko AMADATSU, Hideki KOHNO, Kiminori SATO, ...
    Article type: Article
    1994 Volume 18 Issue 15 Pages 9-14
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    We usually identify people by observing their faces. So, there are many elments of personal identification in the human face. In this report, the result of an experiment on personal identification using right's eye images of 20 persons and right's ear images of 10 persons are shown, and we discuss a possibility of personal identification using part-images consisting of the human face. The way to identify is two kinds of neural network. Finally we got the highest identification rate 95% using eye images.
    Download PDF (709K)
  • Kazuhiro Sanjo
    Article type: Article
    1994 Volume 18 Issue 15 Pages 15-21
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    Recent research in neural network demonstrates the usefulness of neural network for image analysis. This article treats the Landsat multi-spectral image classification using three-layer backpropagation network on the basis of small cluster of pixels. Neural network can be an integrated classifier using both spectral and textural information without explicit measure, however, the uncertainty in presentation procedure of training data makes it difficult to realize robust classification. This report illustrates that the arrangement of training data has significant spatial information around the target pixel and that one can build a robust classifier against multi-category mixture within a single pixel ("mixed pixel" or "mixel") using a neural network learned with various arrangement of training data. Classified result obtained by applying composit-mixel-based training data are compared to the result by conventional pure-pixel-based training data showing better classification for human visual interpretation.
    Download PDF (903K)
  • Moriyuki Matsuo, Takahito Kunimitsu
    Article type: Article
    1994 Volume 18 Issue 15 Pages 23-30
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    This paper describes a new method for image restoration by blind deconvolution using genetic algorithm(GA), named Genetic-Iterative Fourier Transfer Algorithms(GIFTA) . First, they code into gene that are intensity of Fourier transformed two images degraded and initial guess one, and the gene makes an individual with a phase of degraded one. Then, genetic operations, i.e., reproduction, crossover and mutation are done in Fourier domain. Simularity is used in order to estimate the gene in spatial domain. Where, a -priori knowledge which are non-negativity and finitly of an image are constrained. Then, these steps iterate until satisfy the ending condition. Excellent results of simulation are provided by this method.
    Download PDF (788K)
  • Wentao ZHENG, Hiroshi HARASHIMA
    Article type: Article
    1994 Volume 18 Issue 15 Pages 31-38
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    This paper presents a surface representation approach based on 3D invariant characteristics. The basic idea is to represent a surface by dominant points which have important shape attributes. To determine dominant points, we propose a criterion based on 3D invariant (characteristics of surfaces, which we call mean-square curvature. This quantity has some desirable properties and, is suitable for dominant point selection. It also allows of a physical explanation. Using mean-square curvature, we develop an algorithm for dominant point selection. These dominant points are then triangulated in 3D space to construct a wireframe model to represent the surface. Moreover, surface is interpolated based on the model and representation error is shown.
    Download PDF (956K)
  • Naotoshi Fujimoto, Gang Xu, Saburo Tsuji
    Article type: Article
    1994 Volume 18 Issue 15 Pages 39-46
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    Coast line image data are compressed by first dividing the closed contour into segments at high curvature points. For each contour segment, we then compute complexity, which is similar to the fractal dimension, by a function we specially defined for this purpose. Each segment is coded by the coordinates of the two end points, the complexity and a shape pattern that roughly describes the way that the contour deviates from the straight line linking the two end points. In the recontruction stage, stochastical fractal is used to draw the coast line based on the complexity values and local shape patterns so that the resulting contour looks natural and at the same time is close to the original image.
    Download PDF (704K)
  • Toshiyuki Shouhon, Haruo Ogiwara
    Article type: Article
    1994 Volume 18 Issue 15 Pages 47-54
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    A color image is represented by a vector function C^^→=f^^→(x,y) that represents a mapping from a position (x, y) in 2-D plane to a vector C^^→ that represents single color. Taking this fact into account, authors have newly devised a KL transformation for vector function, and applied the transformation to color image coding. The coding method is as follows: (1) Expand a color image f^^→(x,y) in terms of a set of orthonormal vector functions <φ^^→>^<(u)>(x,y) as follows: f^^→(x,y) =Σ_ua_u<φ^^→>^<(u)>(x,y). (2) Transmit the coefficients of the expansion a_u as information of the image. An experiment of the coding method was made on the basis of the RGB color system. The result of the coding method was far better than the JPEG method.
    Download PDF (547K)
  • Ryoma Oami, Masaki Sato, Kiyoharu Aizawa, Mitsutoshi Hatori
    Article type: Article
    1994 Volume 18 Issue 15 Pages 55-62
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    This paper presents a noble motion estimation technique based on single pixel matching. In this proposed method the flexibility of the motion estimation is increased, and also the motion estimation which is adapted to local properties of the image can be easily performed, since the pelwise displaced frame difference (DFD) is calculated in advance. Particularly, the accuracy of the motion estimation along the motion boundary is improved. Several methods based on the single pixel matching are discussed, and the simulation results show the usefulness of the proposed method.
    Download PDF (853K)
  • Akihiko Machizawa
    Article type: Article
    1994 Volume 18 Issue 15 Pages 63-69
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    Interpolative DPCM, though, shows the better prediction performance than extrapolative DPCM, I has a serial problem that the IDPCM has a mount of coding error based on the quantization error. This articl reports an optimization algorithm to decide the offsets and the optimal parameters (subsampling width an quantization width) of an offset technique that adds offsets to pixels to decrease the quantizatione errors. Whe the offset technique is used with subsampling technique. SNR is increased about 3 [dB] and compression rati becomes about 0.3 [bit/pixel] better.
    Download PDF (666K)
  • Mei SUGIHARA, Tsuyoshi HANAMURA, Hideyoshi TOMINAGA
    Article type: Article
    1994 Volume 18 Issue 15 Pages 71-78
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    We propose scalable video coding for t,he scalability function with "bitstream enhancer", to update scalable bitstream outside. The scalability is to provide various size of video image suitable for any decoder from a coded bitstream. In order to solve any problems of scalable video coding for scalability function now, we propose updatable scalability. And one method of updatable scalability, we think about modified simulcast coding added to mode controller between low resolution layer and high resolution layer. As the simulation result, coding efficienty of this coder is almost the same asthat of SNR scalability coder.
    Download PDF (640K)
  • Article type: Appendix
    1994 Volume 18 Issue 15 Pages App1-
    Published: February 25, 1994
    Released on J-STAGE: October 13, 2017
    RESEARCH REPORT / TECHNICAL REPORT FREE ACCESS
    Download PDF (66K)
feedback
Top