logo

SCIENTIA SINICA Informationis, Volume 47, Issue 7: 863(2017) https://doi.org/10.1360/N112016-00057

Color image edge extraction using memristor-based CNN

More info
  • ReceivedSep 3, 2016
  • AcceptedDec 14, 2016
  • PublishedApr 27, 2017

Abstract

Because of the locally connected lattice structure and high-speed parallel processing, cellular neural network (CNN) have been widely used in image processing. Traditional processing methods typically employ fixed templates, which impose significant limitations on practical complex image processing. However, the hardware implementation of large-scale CNNs becomes impossible due to the bottleneck of traditional CMOS technology. In this paper, a new threshold-adaptive algorithm is proposed by considering pixel space distributions based on human visual perception, which can overcome the aforementioned limitation. Then, the memristor, a two-terminal nonlinear device with unique high-speed switching, nonvolatility, and nanometer scale is used to solve the circuit realization problem. Specifically, we design a spintronic memristor-based CNN (SMCNN) to facilitate the proposed threshold-adaptive algorithm. Finally, by using the example of color image processing, the effectiveness of the proposed SMCNN is demonstrated by means of numerical simulations and comparative analysis.


Funded by

国家自然科学基金(61372139,61571372,61101233,60972155)

新世纪优秀人才支持计划(教技函[2013]47号)

教育部“春晖计划"科研项目(Z2011148)

中央高校基本科研业务费专项资金(XDJK2016A001,XDJK2014A009,XD JK2016D008)

留学人员科技活动项目择优资助经费(国家级,优秀类,渝人社办(2012)186号)

重庆市高等学校优秀人才支持计划(渝教人(2011)65号)

  • Figure 1

    Topology of a CNN, in which the squares represent cells with identical structure

  • Figure 2

    (Color onine) The circuit representation of an SM-CNN cell

  • Figure 3

    The structure of a spintronic memristor based on magnetic-domain-wallmotion

  • Figure 4

    (Color online) Characteristic curves of the spintronic memristor. (a) $V$-$I$ Characteristics;(b) threshold characteristics ($V=1.05, f=10$ MHz, parameter of memristor: $D=1000$ nm, $h=7$ nm, $z=10$ nm, Rel = 50, GMR = 12)

  • Figure 5

    (Color online) Original images. (a) Lena; (b) House

  • Figure 6

    The figure of Lena's edge extraction. (a) Sobel; (b) Robert; (c) Prewitt; (d) Log; (e) Canny; (f) color image edge extraction base on self-adaptive threshold (with space factor)

  • Figure 7

    The figure of House's edge extraction. (a) Sobel; (b) Robert; (c) Prewitt; (d) Log; (e) Canny; (f) color image edge extraction base on self-adaptive threshold (with space factor)

  • Figure 9

    The noise polluted figure of Lena's edge extraction. (a) Sobel; (b) Robert; (c) Prewitt; (d) Log; (e) Canny; protectłinebreak (f) color image edge extraction base on self-adaptive threshold (with space factor)

  • Figure 10

    The noise polluted figure of House's edge extraction. (a) Sobel; (b) Robert; (c) Prewitt; (d) Log; (e) Canny; (f) color image edge extraction base on self-adaptive threshold (with space factor)

  • Table 1   Three primary colors vision functions
    Color Wavelength (nm) Luminosity function
    Red 440 0.114
    Green 500 0.587
    Blue 660 0.299
  • Table 2   A comparison of FOM among all kinds of edge detection algorithms
    Test image Sobel (dB) Robert (dB) Prewitt (dB) Log (dB) Canny (dB) Self-adaptive (dB)
    Lena 0.3218 0.3904 0.4371 0.4932 0.5197 0.8492
    House 0.3102 0.3635 0.4062 0.4604 0.4963 0.8177
  • Table 3   A comparison of PNSR among all kinds of edge detection algorithms
    Test image Sobel (dB) Robert (dB) Prewitt (dB) Log (dB) Canny (dB) Self-adaptive (dB)
    Lena 13.6739 13.6466 13.6772 10.0051 8.5541 22.5676
    House 14.2789 14.9443 14.4847 9.7481 7.2392 24.3161

Copyright 2020 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有

京ICP备18024590号-1       京公网安备11010102003388号