【作者】 王廷银；

【导师】 陈荣； 陈冠楠；

【作者基本信息】 福建师范大学 ， 光学工程， 2009， 硕士

【摘要】 双光子显微术因其特别的优点—较低的光漂白与光损伤、较佳的成像对比度以及较深的成像深度,已成为当今生物光学成像的重要技术。本文介绍了双光子成像技术的原理,基于双光子激光扫描共聚焦显微成装置,针对人体鼻咽细胞以及胶原纤维的双光子显微图像,进行了图像处理的实验研究,主要包括鼻咽细胞核分割和鼻咽胶原纤维纹理分析两个部分。在鼻咽细胞核分割中,采用噪声干扰法进行去噪,采用低帽的变换等的数学形态学来增强鼻咽细胞图像,使细胞更加容易分辨。接着对几种经典边缘检测算法进行讨论比较,并根据鼻咽双光子显微图像的实际特征,采取腐蚀算法相结合求出鼻咽细胞边缘。然后进行区域生长定位细胞,采用改进的判别分析算法和区域面积算法对鼻咽细胞进行阈值分割,获得较好结果。对于鼻咽胶原纤维的纹理,首先根据胶原纤维的病理确定出形状、密度、边缘方向等多个可以区分正常鼻咽组织和病变鼻咽组织的特征,而后根据相应特征的提取算法提取各个特征,并用Tamura等算法对提取的特征进行了分析。对福建医科大学附属协和医院提供的10多例双光子显微图像进行分析,结果表明这些特征能够有效地区分正常鼻咽组织和病变鼻咽组织,为医生诊断和治疗此类鼻咽疾病提供科学依据。在文章的最后,作者尝试提出的一种用于图像自动分割的算法。首先通过目标检测算法,确定目标的大致内部和背景区域；然后利用高斯混合模型建立其内部和背景区域的模型；最后迭代地运用图切分算法准确地分割出需要的目标。更多还原

【Abstract】 In recent years, Two-photon Microscope pictures technology has attracted lots of scholars’ interest in the field of biomedicine. That’s because the technology mentioned above obtains several strong points as follows:lower level of light bleach and damage, better contrast ratio of the image and deeper formation of image. The writer tries firstly to probe the basic principle of Two-photon Microscope pictures, and then practices experiment on the pattern process, which contains two parts:the nasopharynx cells division and collagenous thread lamination analysis. The pattern process is carried out on a platform formed by photon laser scanning and focus microscopes, while the objects of the experiments are human nasopharynx cells and Collagenous thread.In this thesis, the writer tries to discuss the method of getting better effect of nasopharynx cells division. The writer holds the opinion that one should adopt the noise denoising method, then a transform of Bottom-hat to enhance the image of nasopharyngeal cells, making the image easier to distinguish; Latter, the writer tries to discuss and compare several classical edge detections. According to the actual characteristics of nasopharyngeal two-photon microscopic image, one can get the edge of nasopharyngeal cells derived by means of corrosion algorithm. Furthermore, one can locate the area of cells, and decide the cells threshold segmentation with the help of improved differentiated analyzing algorithm and regional area algorithm for a better effect.As to collagenous thread lamination, we can firstly figure out the collagenous thread’s form, density, edge direction, etc, which are main features to tell apart nasopharynx organization from pathological changed one, according to pathology of collagenous thread; Then we can draw each characteristic according to corresponding abstraction algorithm of characteristic; Thirdly, analyze the characteristic that is drawn with Tamura algorithm. The result of analyses on more than ten Two-photon Micro- pictures, offered by Affiliated Union Hospital of Fujian Medical University, indicates that these characteristics can effectively distinguish normal nasopharynx organization and pathological changed one, and at the same time can offer scientific basis to doctors in diagnosing and treating this kind of disease.The writer here puts forth his own method tried during the research course in the last part of the thesis. In the author’s eyes, an automatic dividing method can be used in image division. There are three steps in doing this:the first step is to find out the rough inside and background area of the nucleus picture by means of nucleus detection algorithm; the second is to set up a model of inside and background area by utilizing Gauss mixed color model; the third and the last is to derive the target nucleus by using segmentation algorithm in turn.更多还原