基于小波变换和形态学分水岭的血细胞图像分割  

Blood Cell Image Segmentation Based on Wavelet Transform and Morphological Watershed

作　　者：黄籽博 刘任任[1] 梁光明[2,3] 

机构地区：[1]湘潭大学信息工程学院,湖南湘潭411105 [2]国防科技大学电子科学与工程学院,湖南长沙410075 [3]爱威科技股份有限公司研发中心,湖南长沙410013

出　　处：《计算技术与自动化》2017年第3期100-104,共5页Computing Technology and Automation

基　　金：湖南省科技厅资助项目(2016GK4014)

摘　　要：医学图像处理提取细胞中使用分水岭方法时,容易产生过分割现象且对噪声的干扰极为敏感,为了解决此缺点,提出一种基于小波变换和形态学分水岭的细胞图像分割新方法。首先采用小波变换多分辨率分析对图像进行分解,选取合适的小波基和改进去噪阈值函数对图像进行小波去噪,然后对去噪后小波重构的细胞图像应用数学形态学距离变换、灰度重建等技术产生的区域标记进行分水岭变换,最终得到分割结果。实验结果表明,该算法能稳定、准确地提取细胞和实现粘连细胞的自动分割,同时具有很好的鲁棒性和普适性。Method watershed to extract cells during medical image processing will easily caused segmentation and highly sensitive to noise interference.In order to solve this problem,a new method of cell image segmentation based on wavelet transform and Morphological Watershed is proposed.Firstly,decomposing the image by analysis of the wavelet transform multi-resolution analysis to select a suitable wavelet basis and to denoise the image by improved threshold function,then carry out watershed transform labeled region produced by distance transform and gray reconstruction that using mathematical morphological with cell image of wavelet reconstruction after denoising.The final segmentation results will be obtained.The experimental result shows that the algorithm can extract cells stably and accurately and automatically segments adhesion cell,which has strong robustness and adaptation.

关 键 词：小波变换 分水岭 图像分割 粘连细胞 数学形态学 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]