结合小波系数的normalizedcut分割算法.pdf

第 31卷第 1期 计算机应用 Vo.l 31No. 1 2011年 1月 Journal of Comput rApp lications Jan. 2011 : 100 1- 9081( 20 11) 01- 0 182- 02 do:i 10. 3724 /SP. J. 1087. 2011. 00182 normaliz d cut 1, 2 1 黄 颖 , 杨光琼 ( 1. , 400065; 2. , 610054) ( huangy ing@ cqupt. du. cn) : 为了提高 norm aliz d cut分割算法的准确率, 结合小波系数调整图像的亲和矩阵 先采用小波系数计算 图像的边缘信息, 然后使用边缘信息对原始图像构造一个图, 求取图的拉普拉斯矩阵的前 K 个特征值, 并对第二 个特 征值对应的特征向量进行分类, 得到最终的分割结果使用 一些自然图像进行了验证, 结果证明新方法能提取目标更 为细致的细节, 保留更多有用的信息 : 图像分割; normaliz d cut方法; 小波系数; 图分割 : TP39 1. 4 1 : A N orm alized cu t segm en tation algorithm com b ined w ith wavelet coefficien t 1, 2 1 HUANG Y ing , YANG Guangqiong ( 1. College of Comp uter Science and Technology, Chongqing University of Posts and Telecommunications, Chongqing 400065, China; 2. School of E lectronic Eng ineering, University of E lectronic Science and Technology of China, Chengdu Sichuan 610054, China) A bstract: W av l t co ffici ntw as us d to adjust th affin itym atrix of th mi ag and to g tm or accurat r su lts. Firstly, th dg in form ation of th mi ag w as calcu lat d w ith w av l t co ffici nt. S cond ly, a graph w as construct d for th original mi ag w ith dg inform ation, and th Lap lacian m atrix was obtain d and th firstK ig nvalu s w as comput d. Th ig nv ctor corr spond ing to th s cond ig nvalu was classifi d to g t th final s gm ntation r sults. Th xp rmi ntal r su lts show that th propos d m thod can xtract mor fin d tails, and