Design of Nonlinear Cellular Neural Network Filters for Detecting Linear Trajectory Signals

• MUIKAICHI Masahiro

  Faculty of Science and Technology, Keio University

• KONDO Katsuya

  Faculty of Science and Technology, Keio University

• HAMADA Nozomu

  Faculty of Science and Technology, Keio University

抄録

Recently, the spatio-temporal filter using linear analog Cellular Neural Network (CNN), called CNN filter array, has been proposed for the purpose of dynamic image processing. In this paper, we propose a design method of discrete-time cellular neural network filter which selectively extracts the particular moving object from other moving objects and noise. The CNN filter array forms a spatio-temporal filter by arranging cells with a same function. Each of these cells is a simple linear analog temporal filter whose input is the weighted sum of its neighborhood inputs and outputs and each cell corresponds to each pixel. The CNN filter is formed by new model of discrete time CNN, and the filter parameters are determined by applying backpropagation algorithm in place of the analytic method. Since the number of connections between neurons in the CNN-type filter is relatively few, the required computation in the learning phase is reasonable amount. Further, the output S/N ratio is improved by introducing nonlinear element. That is, if the ratio of output to input is smaller than a certain value, the output signal is treated as a noise signal and ought to be rejected. Through some examples, it is shown that the target object is enhanced in the noisy environment.

収録刊行物

• IEICE transactions on fundamentals of electronics, communications and computer sciences

  IEICE transactions on fundamentals of electronics, communications and computer sciences 80 (9), 1655-1661, 1997-09-25

  一般社団法人電子情報通信学会

キーワード

• cellular neural network
• dynamic image processing
• extraction of moving object
• linear trajectory signal

詳細情報

• CRID: 1571417127334127232
• NII論文ID: 110003207853
• NII書誌ID: AA10826239
• ISSN: 09168508
• 本文言語コード: en
• データソース種別:
  • CiNii Articles