工业检测图像灰度波动变换自适应阈值分割算法  

An Adaptive Thresholding Algorithm Based on Grayscale Wave Transformation for Industrial Inspection Images

作　　者：魏巍[1,2] 申铉京[1,2] 千庆姬[3] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012 [2]吉林大学符号计算与知识工程教育部重点实验室,长春130012 [3]吉林大学物理学院,长春130025

出　　处：《自动化学报》2011年第8期944-953,共10页Acta Automatica Sinica

基　　金：国家自然科学基金(60773098);吉林省科技发展计划(20080317)资助~~

摘　　要：工业检测图像经常受到不均光照的影响,对该类图像局部自适应分割算法比全局算法能产生更好的分割效果.但局部算法中基于分块的算法对分块方法缺乏指导,而基于邻域的算法容易在背景或前景内部产生误分.针对上述缺点,本文提出了一种多方向灰度波动变换的自适应阈值分割算法.该算法先从多个方向依照灰度波动对图像进行转换,构造以多维向量为基础的灰度波动变换矩阵,然后利用主成分分析法(Principal component analysis,PCA)将高维向量压缩至一维并生成变换图像,最后运用Otsu算法分割变换图像.该算法无需分块,并且仅需波动幅度阈值和布尔型背景色两个参数.实验结果表明,该算法能够有效减少不均光照对工业检测图像分割的影响,与Niblack法、Sauvola法等几种局部算法相比,该法在分割效果上具有了明显的提升.The industrial inspection images are usually under non-uniform illumination, and local adaptive thresholding algorithms have better segmentation performance on them than the global ones. But the local algorithms based on image＇s sub-blocks are short of instructions for partitioning, and the local algorithms based on pixel＇s neighborhood will probably cause some misclassifications within the background or foreground. To resolve these problems, a novel adaptive thresholding algorithm based on multi-directional grayscale wave transformation is proposed in this paper. Firstly, it performs the transformation by grayscale waves in multi-directions to get a matrix of multi-dimensional vectors. Secondly, the vectors are compressed to one dimension using the principal component analysis （PCA） method, and then the Otsu global method is employed to find optimal wave threshold for segmentation on this matrix. This algorithm does not need partitioning the image any more and only takes the peak height threshold and the boolean background color as its two parameters. Experiments demonstrate that this method has a excellent capability of decreasing the influence of non-uniform illumination in industrial inspection images, and its segmentation performance is better than several other local thresholding algorithms, such as Niblack＇s method and Sauvola＇s method.

关 键 词：图像分割 局部阈值分割 灰度波动 OTSU算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]