基于神经网络和模糊算法图像边缘检测方法.docVIP

基于神经网络和模糊算法图像边缘检测方法 　　 (燕山大学 信息科学与工程学院， 河北 秦皇岛 066004) 　　?? 　　摘要：针对传统图像边缘检测方法中出现毛边、噪边、边缘定位不精确等缺点，提出一种神经网络与模糊算法相结合的检测方法。根据图像特征，将图像分为高频和低频部分分别处理，高频部分适宜用双层网络结构，可以很好地减弱噪声；对于图像低频部分，将模糊理论引入到边缘检测中，能够检测出弱边。最后对检测出的两个图像边缘进行融合，实验结果证明得出的检测效果较好，比传统边缘检测算子所获结果有很大改善。 ?? 　　关键词：边缘检测； 神经网络； 模糊算法； 隶属度函数?? 　　中图分类号：TP391文献标志码：A 　　文章编号：1001-?B3695(2008)11-?B3507-?B02 　　?? 　　Image edge detection based on neural network and fuzzy algorithm 　　?? 　　JIA Chao, ZOU Qi, YAO Fang, WANG Bei-bei, AI Dong 　　??(College of Information Science Engineering, Yanshan University, Qinhuangdao Hebei 066004, China) 　　??Abstract:Considering the disadvantages in the traditional image edge detection methods, such as rough edge, noise edge and inaccurate edge location,this paper proposed a novel edge detection approach which was the combination of neural network with fuzzy algorithm to solve above problems. First, according to the character of image, the image was divided into high frequency and low frequency, and processed respectively. It applied the method of LVQ (learning vector quantization) double-layer neural network to the high frequency part of image. The noise could be suppressed effectively with this method. Then using the fuzzy algorithm for the low frequency part of image, thus weak edges could be detected with high quality, and the target could be separated from the background clearly. Finally, combined the two images from former methods. The experimental result demonstrates that the effect of the edge detection is greatly improved comparing with the traditional edge detection methods.?? 　　Key words:edge detection; neural network; fuzzy algorithm; membership function?お? 　　边缘检测[1,2]是图像处理和计算机视觉等领域最基本的技术，广泛用于轮廓抽取、特征检测和纹理分析。人们在这方面已作了不少研究，且提出一些算法，经典的边缘检测算子有Sobel算子、Prewitt算子、Kirsch算子、Roberts算子等[3]，这些算法虽然计算简单，但对噪声较敏感。本文首先用小波变换对图像分频，然后用神经网络和模糊算法对图像的不同频段分别处理。图像的高频部分用神经网络方法检测，忽略低频分量包含的部分边缘信息；而图像的低频分量边缘比较模糊，将模糊集理论引入到图像边缘检测方法中，能够有效地将目标从背景中分离出来，达到较好效果；然后对分离出的边缘图像运用差影法进行融合。两种方法检测出的图像边缘具有良好的互补性，图像去噪