微型机与应用
微型機與應用
미형궤여응용
MICROCOMPUTER & ITS APPLICATIONS
2012年
16期
42-45
,共4页
彩色图像分割%差分演化%Neut准则%爬山法
綵色圖像分割%差分縯化%Neut準則%爬山法
채색도상분할%차분연화%Neut준칙%파산법
color image segmentation%differential evolutional%normalized cut%hill climbing algorithm
针对解Ncut准则的SM算法寻优能力不足的问题,提出一种基于差分演化优化归一化准则的彩色图像分割算法。首先对彩色图像进行爬山法预分割为多类,并构造类级间的无向完全图,之后再使用二进制差分演化算法求得Ncut准则最小化的图二分,最后通过映射获得图像的二值分割。实验结果表明,在相同预处理情况下,本文的寻优算法与SM算法相比,分割效果更为精准。
針對解Ncut準則的SM算法尋優能力不足的問題,提齣一種基于差分縯化優化歸一化準則的綵色圖像分割算法。首先對綵色圖像進行爬山法預分割為多類,併構造類級間的無嚮完全圖,之後再使用二進製差分縯化算法求得Ncut準則最小化的圖二分,最後通過映射穫得圖像的二值分割。實驗結果錶明,在相同預處理情況下,本文的尋優算法與SM算法相比,分割效果更為精準。
침대해Ncut준칙적SM산법심우능력불족적문제,제출일충기우차분연화우화귀일화준칙적채색도상분할산법。수선대채색도상진행파산법예분할위다류,병구조류급간적무향완전도,지후재사용이진제차분연화산법구득Ncut준칙최소화적도이분,최후통과영사획득도상적이치분할。실험결과표명,재상동예처리정황하,본문적심우산법여SM산법상비,분할효과경위정준。
To overcome the inadequacy of ability of SM algorithm in solving Ncut optimal value, this paper proposed a color image segmentation algorithm based on differential evolution algorithm. Firstly, color image is clustered by hill climbing algorithm to construct an class-level undirected complete graph. Then, it gets an bipartite graph with least Ncut value based on binary differen- tial evolution algorithm. Finally, it obtains the image's binary segmentation by mapping. The experiments show that this method obtains more precise color image segmentation resuh compared with the SM algorithm in the same pretreatment case.