An Applied Research Of Pulse Coupled Neural Network For Image Processing

Pulse coupled Neural Network (PCNN) is a new type of Neural Network. In this paper, asimplified pulse coupled neural network is improved, and is applied to image processing,mainly from two aspects, the image denoising and image segmentation in the coal mine imageprocessing.Aims at the traditional image denoising methods make the image blur and the image edgeinformation loss and so on, due to darkness, uneven luminosity and large noise ofunderground coal mines, according to the characteristics of underground mine images, anintegrated algorithm of salt and pepper noise reduction for underground mine image based onmodified simplified PCNN model was proposed. This algorithm modifies neurons joiningstrength.Improved neurons joining strength depends on the pixel gray value of noiseimage, thus, the treatment effect of salt and pepper noise reduction is better. At the same time,this algorithm also improves the decay time constant Eof dynamic threshold, whichdepends on the normalization constantv E. Number of parameters in the model is reduced.Then, the value of normalization constant is selected through the experiment withexperience. Experiment results indicate that this modified simplified PCNN model canremove salt and pepper noise of underground mine images and preserve the details of imagesmore effectively and completely than the conventional methods.Improved simplified pulse coupled neural network with the minimum cross entropy isapplied to image segmentation. Threshold attenuation functionE ijof simplified pulsecoupled neural network is Improved, and it depends on. Other parameters are the same asthe image denoising method of parameter selection. Experiment results indicate that Improvedsimplified pulse coupled neural network with the minimum cross entropy is better than thetraditional threshold segmentation method for image segmentation, because they can segment the more edge details.