结合离散熵和自组织神经网络_SOM_的边缘检测方法.pdfVIP

第29 卷第5期 东 北 大 学 学 报 ( 自 然 科 学 版 ) Vol29, No. 5 2 00 8 年 5 月 Journal of Northeastern U niversit ( Natural Science) M a 2 0 0 8 (SOM) 王 坤, 高立群, 片兆宇, 郭 丽 ( , 110004) : , , ; ; 6 6 , SOM , : ; ; ; ; : TP 391 : A : 2008) 05064 104 Edge Detection Method Combing Discrete Information Entropy with SelfOrganizing Map WAN G K un, GA O L iqun, PIAN Zhaoyu, G UO L i ( School of Information Science Engineering , Northeastern Universit , Shen ang 110004, China. Correspondent: WA NG Kun, Email: ogo w @ 163. com) - Abstract: An edge detection method is proposed combining image discrete information entrop w ith selforganizing map ( SOM) . A threshold is chosen from some different information entropies to segment the smooth region from the region w here the gra level abruptl changes so as to reduce computation. T hen, the gra level images are transformed into the ideal binar pattern of pixels. Six t pes of edge and six protot pe vectors are defined, among them the latter are taken as inputs into SOM to classif the edge t pes and then obtain edge images from which the speckled edges are removed to acquire ideal edge images. Experimental results show ed that the edge images gained b the method proposed are better than those b other edge detection methods. Key words: edge detection; discrete information entrop ; threshold; selforganizing map; speckle noise , , , [ 1] , Sobel Prew itt Robert ,