基因芯片图像高亮噪声处理算法研究  

Methods for Highly Fluorescent Noise Reduction in Microarray Images

作　　者：邵桂芳[1] 李刚[2] 李铁军[3] 张倩[1] 

机构地区：[1]厦门大学模式识别与智能系统研究所,福建厦门361005 [2]重庆工学院计算机学院,重庆400050 [3]集美大学信息工程学院,福建厦门361021

出　　处：《厦门大学学报（自然科学版）》2009年第3期330-333,共4页Journal of Xiamen University：Natural Science

基　　金：国家自然科学基金(60443004);重庆市自然科学基金(2007BB2415)资助

摘　　要：基因芯片图像噪声处理决定了后续处理与分析的精度和准确性.高亮斑点噪声是一种亮度很高、成块状的特殊噪声,应用传统噪声处理方法无法滤除.针对高亮噪声的特点,首先利用形态学碟形结构元素进行开运算来评价图像是否含有噪声,然后利用阈值分割去除噪声.提出两种阈值获取方法,一种是基于最大类间方差的自适应获取方法,一种是固定阈值法.通过大量定性和定量实验比较分析,结果表明,两种方法都可以有效除高亮斑点噪声.Microarray image noise reduction determines the precision and veracity of microarray data processing and information extraction. Highly fluorescent noise is a type of special noise with higher brightness and shape in block which can＇t be removed by traditional methods. According to the feature of it, this paper proposes that using mathematical morphology opening operator with disk structure element to evaluate image quality, then suppress the noise with thresholding segmentation techniques. Two methods are put forward to get threshold：one is the adaptive method based on maximum variance between clusters, another is the fixed means. Through comparison and analysis, the lots of experiment results show that both two methods can effectively remove the highly fluorescent noise and preserve edge and detailed information.

关 键 词：基因芯片 去噪 数学形态学 最大类间方差 阈值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]