节点文献

动物骨髓细胞图像分割方法的研究

Studies on Segmentation Methods of Animal Marrow Cells Image

【作者】 侯振杰

【导师】 麻硕士;

【作者基本信息】 内蒙古农业大学 , 农业机械化工程, 2005, 博士

【摘要】 针对在动物医学领域中对动物骨髓病理图像的分析与识别仍属空白这一现状,并考虑到图像分割是进行正确分类和识别的基础,是图像处理领域中的关键问题之一,因此利用不同的理论和算法对动物骨髓细胞图像进行了分割方法的研究。 针对图像采集过程中的噪声问题,提出了一种改进的中值滤波算法。 以空间尺度滤波为基础,提出了一种多阈值的细胞分割方法,解决了骨髓细胞切片图像直方图没有明显峰谷值从而导致阈值选取困难的问题。将小波分析与模糊C 聚类分析相结合,从而完成更细致的图像分割任务。利用二维连续 GAUSS 小波在不同方向不同尺度对图像进行检测并做加权处理,达到了对具有方向性图像进行分割的目的。 利用腐蚀与膨胀等算法,提出了一种适合于骨髓细胞图像分割的算法。针对分水岭算法产生的过度分割现象,结合一个模板进行了算法的改进。针对多尺度梯度检测算子所获得的细胞轮廓曲线不连续等问题,利用活动轮廓模型进行了优化,获得了连续闭合曲线。 研究了基于分形学的动物骨髓细胞图像分割方法。利用一个简单的计算方法来近似代替 HOLDER 指数并通过计算其多重分形谱以寻求合适的分割阈值。 在标准遗传算法的基础上,结合聚类分析,将信息学中的熵作为适应度函数,提出了一种基于熵的聚类遗传算法并应用于骨髓细胞图像分割中。

【Abstract】 Aiming at the fact that there are no existing methods of analysis and recognition for animal marrow pathology image in animal medical field, and segmentation is one of the key procedures in digital image processing field and the basis of correct classification and recognition, this paper has a research on segmentation methods of animal marrow cells image according to various theories and algorithms. Produces an improved median filter algorithm for better denoising. Provides a multi-threshold cell segmentation method for solving the difficult problems choosing the threshold because there is no obvious apex and valley value in marrow cell slice image histogram based on spacial scale filter. This paper also integrates the wavelet analysis and blurring C clustering analysis. At the same time, we can segment the directional image by detecting and adding priority in different direction and scale using two-dimension continuous GAUSS wavelet transform. By adopting erosion and dilation algorithm, this paper provides a good algorithm for marrow cells image segmentation. Aiming at the over-segmentation phenomenon produced by using watershed algorithm, this paper improves this algorithm in virtue of a panel. We can get the continuous loop curves by taking active profile model into withdrawing nucleolus profile curves according to the discontinuous cell profile curves obtained by using multi-scale grade operator. Does a research on segmentation methods of animal marrow cells image based on morphology. We can replace the HOLDER exponent by a simple calculation method and find a suitable segmentation threshold by calculating its multi-fractal spectrum. Provides a genetic clustering algorithm on entropy and applies it into marrow cells image segmentation by taking entropy in information science as fitness function based on standard genetic algorithm and clustering analysis.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络