Stochastic Resonance Based Visual Perception Using Spiking Neural Networks

Yuxuan Fu; Yanmei Kang; Guanrong Chen

doi:10.3389/fncom.2020.00024

Stochastic Resonance Based Visual Perception Using Spiking Neural Networks

Front Comput Neurosci. 2020 May 15:14:24. doi: 10.3389/fncom.2020.00024. eCollection 2020.

Authors

Yuxuan Fu¹, Yanmei Kang¹, Guanrong Chen²

Affiliations

¹ Department of Applied Mathematics, School of Mathematics and Statistics, Xi'an Jiaotong University, Xi'an, China.
² Department of Electrical Engineering, City University of Hong Kong, Hong Kong, China.

Abstract

Our aim is to propose an efficient algorithm for enhancing the contrast of dark images based on the principle of stochastic resonance in a global feedback spiking network of integrate-and-fire neurons. By linear approximation and direct simulation, we disclose the dependence of the peak signal-to-noise ratio on the spiking threshold and the feedback coupling strength. Based on this theoretical analysis, we then develop a dynamical system algorithm for enhancing dark images. In the new algorithm, an explicit formula is given on how to choose a suitable spiking threshold for the images to be enhanced, and a more effective quantifying index, the variance of image, is used to replace the commonly used measure. Numerical tests verify the efficiency of the new algorithm. The investigation provides a good example for the application of stochastic resonance, and it might be useful for explaining the biophysical mechanism behind visual perception.

Keywords: contrast enhancement; spiking networks; stochastic resonance; variance of image; visual perception.