基于支持向量机的血液细胞核彩色图像分割

Color Segmentation of Nuclei of Blood Cell Using Support Vector Machines

对血液细胞核进行精确的分割是自动分析与识别的关键环节,现有经典算法很难获得满意的效果。本文将分割问题转化为分类问题,利用支持向量机(SVM)实现血液细胞核彩色图像分割。为了获得最佳的分割效果,对采用不同色彩空间、核函数及样本数量的分割结果进行了详细的比较和分析。实验结果表明,与目前经典的分割算法比较,该算法具有分割速度快、准确率高及泛化性强等优点。

To extract nuclei of white blood cell from the micro-image in visual field with microscope,a novel color image segmentation scheme using Support Vector Machines（SVM） is proposed. In this paper,the SVM is trained on 4097 positive samples made up of white blood cell nuclei pixels and 2910 negative samples（some red blood cells and background pixels）. To improve the performance of real images, detailed analysis and comparisons are made by choosing different color spaces,kernel functions,samples. We demonstrate experimentally that the proposed algorithm achieves better performance than the existing methods, while being more rapid, accurate and robust.