基于视觉感知和等周割的二维阈值分割算法  

Image Thresholding Segmentation Based on Visual Perception and Isoperimetric Cut

作　　者：邹小林[1] 冯国灿[2] 

机构地区：[1]肇庆学院数学与信息科学学院肇庆526061 [2]中山大学数学与计算科学学院广州510275

出　　处：《计算机科学》2013年第6期303-307,共5页Computer Science

基　　金：国家自然科学基金项目(60975083,61272338)资助

摘　　要：二维阈值分割方法没有考虑人类视觉感知的特性,将整个灰度级区域作为分割阈值的搜索空间。同时等周割图像分割方法没有直接考虑图像的灰度信息以及迭代终止条件难以确定的问题,因而对灰度图像的分割效果不甚理想。因此提出了一种融合视觉感知和等周割的二维阈值分割方法,该方法首先利用视觉感知的特性选择候选阈值向量所在的灰度区域,再将等周割作为准则,从候选阈值向量中选出最小等周率所对应的候选阈值向量作为最佳的分割阈值向量。在一系列图像上的实验结果表明,与几种经典的二维阈值分割方法相比,所提算法的分割效果更好。Because the two-dimensional threshold segmentation methods do not consider the characteristic of human vi- sual perception, the search space of these methods is the whole two dimensional gray-level area. At the same time, image segmentation based on isoperimetric cut does not consider the gray intensity and it＇s iteration terminate condition is dif- ficult to determine, so the effect of image segmentation result using the method is not ideal. In this paper, a novel two- dimensional thresholding method based on visual perception and isoperimetric cut was presented. The proposed method first utilizes characteristics of visual perception to find a gray level vector decided by candidate threshold vectors, then u- ses isoperimetric cut as a criterion to select the minimum isoperimetricratio corresponding to the candidate threshold vector as the optimal threshold vector from the candidate threshold vectors. Experimental results on a series of image show that the proposed method outperforms some classic two-dimensional thresholding methods hasegmentafion quality.

关 键 词：图像阈值分割 等周割 阈值向量选择 视觉感知 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]