摘要
线性轮廓信息有效识别在分层像素级图像融合中具有重要应用。目前常见的轮廓信息识别方法一般基于图像灰度的变化来判断分层像素级图像中的点是否属于轮廓点,识别到的点可能不连续,识别过程导致了融合后图像优质系数的降低、耗时长。针对这类问题,提出一种基于图像分割的分层像素级图像融合中线性轮廓信息识别方法。在分层像素级图像融合过程中,采用多阈值Otsu方法分割分层像素级图像,提取各层轮廓并计算轮廓链的局部转角比率和逐点向量积,描述基于逐点比较法的图像细节处的轮廓特征,结合分层像素级图像结构的经验判据对分层基础数据进行检测和推断,实现分层轮廓点的初步识别,通过不同层次的轮廓像素拟合完成整体图像轮廓的初步定位,在分层像素级图像多特征检测的基础上进行图像信息融合,结合自适应阈值识别的各层次像素轮廓位置检测对整体线性轮廓定位信息进行修正以提高线性轮廓信息识别的准确性。实验结果表明,所提方法有效提高了连续像素边缘比以及清晰度,识别所花时间较少。
This article proposes an information identification method for linear profile in fusion of hierarchy pixel - level image based on image segmentation. Our research used multi - threshold Otsu method to divide the hierarchy pixel - level image during process of the fusion and described contour feature of image detail based on point - by - point comparison method, then detected and inferred hierarchy basic data integrated with experience criterion of structure of the hierarchy pixel - level image to achieve primarily identification for hierarchy contour point. The research completed primary location for whole picture contour via fitting of contour pixel of different levels. Based on multi - feature detection of the hierarchy pixel - level image, the research carried out image information fusion. The research amended the overall linear profile location information to improve identification accuracy integrated with loca- tion detection of each hierarchy pixel profile identified by adaptive threshold. Simulation results show that the method can improve ratio contour and definition of continuous oixel. Identification time is less than traditional method.
作者
宋薇
SONG Wei(Sichuan Technology and Business University, Chengdu University of Information Technology, Chengdu Sichuan 611745, China)
出处
《计算机仿真》
北大核心
2018年第6期408-411,431,共5页
Computer Simulation
关键词
分层
像素级图像
图像融合
线性轮廓
轮廓信息识别
Layered
Pixel -level image
Image fusion
Linear contour
Contour information identification
