噪声在脉冲耦合神经网络图像增强中的作用

Enhancement of Digital Image by Pulse Coupled Neural Networks with Noise

神经系统中广泛存在着噪声,大量研究表明噪声有助于弱信号的检测和传输。脉冲耦合神经网络是建立在生物神经系统上的第三代人工神经网络,被广泛应用于图像处理。为了研究噪声对脉冲耦合神经网络图像处理的影响,通过在网络中引入加性噪声,用于图像增强。直观视觉效果和图像直方图均表明适当的噪声有助于图像增强,噪声过小或过强则减弱图像增强效果;图像的峰值信噪比随噪声强度增强呈现倒钟形,表明存在随机共振现象。本研究表明适当强度的噪声能够提高脉冲耦合神经网络图像处理的效果,并显示出随机共振,有助于开展基于生物神经系统的智能化图像处理方法的研究。

Noise has been found in many kinds of neural systems and thought to be helpful in detecting and processing weak signals. The pulse coupled neural network （PCNN） is a kind of artificial neural network based on the biological experiments, and has been widely applied to image processing. To investigate the influence of noise on the image processing by PCNN, we added a noise to the net to realize image enhancement. The images and their histograms showed that noise of suitable intensity was helpful to the image enhancement, while the noise with smaller or stronger intensities produced harmful effects on the results. The curve of peak signal-to-noise ratio （PSNR） versus noise intensity was reversed-bell like, representing the characteristic of stochastic resonance. It was showed that noise could improve the image enhancement by PCNN through stochastic resonance, being beneficial to the neural system for realizing intelligent signal processing.