脉冲耦合神经网络在图像处理中的应用研究

发布时间：2018-01-17 19:25

本文关键词：脉冲耦合神经网络在图像处理中的应用研究　出处：《兰州大学》2003年硕士论文　论文类型：学位论文

【摘要】：通过对猫等小型哺乳动物视觉皮层的研究而建立的Eckhorn[1]模型是脉冲耦合神经网络(Pulse Coupled Neural Network,PCNN)的基础,经由Johnson和其他研究者[2,3,4,5,6]做了进一步变形而最终演化为PCNN。PCNN 不同于传统的人工神经网络模型,它是通过模拟视觉皮层神经细胞的活动而建立的神经网络模型,是对真实神经元的简化与近似。PCNN网络模型所具有的连接域特性和动态阈值衰减特性能够使状态相似的神经元同步输出脉冲,这一点充分模拟了哺乳动物视觉皮层神经元的生物特性,因而在图像分割、边缘提取、目标识别等图像处理方面获得了广泛的应用。　同时,PCNN 是一种多参数神经网络模型,其应用效果的好坏在很大程度上取决于参数的设置,但是到目前为止各参数对网络模型的影响只存在定性分析,还没有一种针对不同图像从而自动求得各参数最佳值的算法,只能通过手动实验分析来逐步设置。因此最佳参数的设置就成了一项繁琐但又十分关键的工作。本论文对原始PCNN 模型作了一定程度的简化,在保持PCNN 连接域特性和动态阈值衰减特性的基础上减少了神经元模型输入域和连接域的一些参数,并将该简化模型应用到图像降噪和分割编码中,对其实验结果进行了分析和总结。　本论文第二章对该简化模型在滤除图像中的脉冲噪声和高斯噪声方面的应用做了初步探索,在结合前人研究的基础上结合局部中值算法,提出了基于脉冲耦合神经网络的脉冲噪声滤波器和高斯噪声滤波器,并和其他算法相比较,进行了分析和总结。本论文第三章分析了脉冲耦合神经网络在图像分割方面的应用,并将该简化神经网络模型作为分割图像编码的分割算法,应用到图像编码中,提出了基于脉冲耦合神经网络的分割图像编码,最后和基于块变换的编码方法进行了初步比较,分析了各自的优缺点。　
[Abstract]:Eckhorn based on a study of the visual cortex of small mammals such as cats. [Model is the basis of Pulse Coupled Neural Network PCNN, via Johnson and other researchers. [The PCNN.PCNN model is different from the traditional artificial neural network model. It is a neural network model established by simulating the activities of nerve cells in the visual cortex. The simplification and approximation of real neurons. PCNN network model has the characteristics of connection domain and dynamic threshold attenuation, which can synchronize the output pulses of neurons with similar states. This fully simulates the biological characteristics of mammalian visual cortical neurons and has been widely used in image processing such as image segmentation edge extraction target recognition and so on. At the same time, PCNN is a multi-parameter neural network model, the effect of its application depends on the setting of parameters to a great extent, but so far, there is only qualitative analysis of the influence of each parameter on the network model. There is no algorithm for finding the optimal value of each parameter automatically for different images. Only through manual experimental analysis to set up step by step, so the setting of the best parameters has become a tedious but very important work. This paper simplifies the original PCNN model to a certain extent. On the basis of preserving the characteristics of PCNN connection domain and dynamic threshold attenuation, some parameters of input domain and connection domain of neuron model are reduced, and the simplified model is applied to image denoising and segmentation coding. The experimental results are analyzed and summarized. In the second chapter of this paper, the application of the simplified model in filtering impulse noise and Gao Si noise in the image is preliminarily explored, and the local median algorithm is combined with previous studies. The impulse noise filter and Gao Si noise filter based on impulse coupled neural network are proposed and compared with other algorithms. In the third chapter of this paper, the application of impulse coupled neural network in image segmentation is analyzed, and the simplified neural network model is applied to image coding as the segmentation algorithm of segmentation image coding. A segmented image coding method based on impulsive coupled neural network is proposed. Finally, a preliminary comparison is made with the coding method based on block transform, and their advantages and disadvantages are analyzed.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2003
【分类号】：TN911.73

【引证文献】