超微弱生物发光图像的野值分析方法提取  

Extraction of the Ultra-Weak Luminescence Bio-Images Using Outlier Analysis

作　　者：蔡利栋[1] 傅瑜[1] 李锷[1] 张红军[1] 

机构地区：[1]暨南大学计算机科学系,广东广州510632

出　　处：《计算机工程与科学》2009年第9期42-44,共3页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60275028)

摘　　要：生物体可以应激而发射光子生成超微弱发光图像。这类图像的提取是一个超低信噪比信号处理的问题。本文提出可把超微弱发光图像视为服从高斯分布的背景噪声的影像,其部分像素的灰度因受生物信号的不规则干扰而异常偏离分布均值。于是,只需对单幅超微弱发光图进行野值分析就可实现不同信噪比的信号图像提取。实验表明,这一方法能快速、有效地给出图像提取的初步结果,作为进一步的形态学或区域生长处理的"种子"。laving systems may emit photons in response to stress and generate an ultra-weak luminescence bio-image. Acquisition of such images is a problem of extracting the signals of ultra-weak S/N ratio. In this paper, the ultra-weak luminescence image is seen as the image of a Gaussian background noise where a part of pixels have abnormal grey deviations from the mean of the Gaussian due to irregular perturbation from the bio-signals. Therefore, image extraction for bio-signals of different S/N ratios can be made using an outlier analysis on a single ultra-weak luminescence image. The experimental results show that it is an expeditious approach to extracting a preliminary image as the ＂seed＂ for further morphological or region-growing processing.

关 键 词：超微弱发光 野值分析 中位数 稳健方差估计 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]