【作者】 刘颖；

【导师】 阿木古楞；

【作者基本信息】 内蒙古农业大学 ， 计算机应用技术， 2013， 硕士

【摘要】 在图像处理技术用于虫卵图像处理时，受到诸多因素的影响。人工分辨缺乏客观性、稳定性和精确性，处理能力完全取决于研究人员的经验，并且得到的标本、数据和结果不便于存储、再现和查找，所以这种处理方式不能够很好地适应现代信息化处理发展的要求。生物医学图像处理同样随之发展成为这一技术领域十分重要的一个部分。图像处理的核心技术是图像分割，然而相异图像的分割方法没有统一的方法可寻，图像分割目前仍是一个十分热门研究的问题。本文针对虫卵图像分割方法进行研究，寻找适合这种图像的分割方法，最终实现虫卵图像的有效分割。1.分析虫卵图像特性，针对虫卵二值图像设计形态学滤波器，消除杂质及噪声的影响。为后续处理打下基础。2.认真学习分水岭算法基本原理，针对虫卵图像单粘连特性，提出了基于边界脊线标记的虫卵图像分割方法。在分水岭算法和数学形态学处理基础上，提取粘连边界脊线标记进而能够快速准确分割单粘连虫卵图像。实验表明，此方法对单粘连虫卵图像较为适用。3.针对目前形态学分水岭变换的过分割问题，融合多方法将影响分水岭分割过程的区域极小值合并，有效避免过分割现象，取得很好的效果。实验表明，该算法能够有效分割粘连虫卵图像，对多粘连虫卵图像较为适用并且有效地避免了过分割现象。为虫卵图像接下来所要研究的的计数及识别奠定基础。更多还原

【Abstract】 The image processing technology influenced by many factors, such as resolutionlack of objectivity, stability and accuracy, when egg images processing. On the otherhand, processing performance depends entirely on the researcher’s experience, and theobtained samples, data and results are not easy to store, rebuilt and search, so this methodcan not adapt well to the demands of the development of the modern informationprocessing. The key issue of image processing is image segmentation, but differentimage segmentation method has no unified method, so image segmentation is still a verypopular research task now.Aiming at the egg image segmentation, we study some suitable new method for eggimages partitioning, and finally completed the effective segmentation of egg images inthis paper.1. Through analysis on the characteristics of Egg images, morphological filter, toeliminate the influence of impurity and noise, were designed for binary egg images. Sowe lay a good foundation for the subsequent processing.2. Carefully studying the basic principles of watershed algorithm, egg imagessegmentation method is proposed based on the boundary ridge line marker. Based on thewatershed algorithm and mathematical morphology，we proposed boundary ridge linemarker which can fast and accurate segmentation of single adhesive egg images.Experiments show that, this method suitable for simple adhesion egg images.3. In view of the present watershed transform method, multiple methods fusion wereproposed to solve over-segmentation by minimum value combination, and effectivelyavoid over-segmentation, achieved ideal results. Experiment shows that the algorithmcan effectively segment complicated adhesion egg images, and effectively avoidover-segmentation phenomenon. This work provided a good basis for counting andidentification of egg image for the further study.更多还原