根据特征点及梯度进行改进的矩不变阈值分割图像算法  

An Algorithm of Threshold Image with Invariable Intensity Moment Improved by Points of Interest and the Gradient

作　　者：林定天[1] 蔡光程[1] 

机构地区：[1]昆明理工大学理学院,云南昆明650093

出　　处：《昆明冶金高等专科学校学报》2006年第5期23-26,共4页Journal of Kunming Metallurgy College

基　　金：国家自然科学基金项目:格上动力系统的固步模式混沌化问题(10561004)

摘　　要：介绍了图像的矩以及在保持图像的矩不变的条件下对图像进行两级阈值分割的方法,但该算法具有一定的盲目性,分割出来的图像具有过多的无用背景信息。为此,在矩不变阈值算法的基础上提出了一种改进算法,该算法根据图像的特征点及其梯度进行调整,以达到减少盲目性,尽量保留图像前景物体信息,去掉无用背景信息的目的。The intensity moment of images and the method used for the bi-level threshold of images with invariable intensity moment are introduced. The algorithm, however, has some blindness and the images segmented in this method may have much useless background information, therefore, an improved algorithm based on this method is suggested. The algorithm is to adjusted according to the points of interest and the threshold value of the gradient to decrease blindness, hold useful foreground information and get rid of useless background information.

关 键 词：图像分割 阈值 矩不变 特征点 梯度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]