基于特征块的细胞核提取方法

Extraction of Nuclear Regions from Blood Microscopic Images Based on Feature Blocks

提取细胞核图像是定量分析有核细胞形态的重要前提 .该文针对血液显微图像中细胞核的颜色不均匀性 ,仿照人的视觉以整体的综合特征为主导的识别原理 ,提出了一种基于特征块的细胞核自适应提取方法 .首先寻找条件比较典型的细胞核特征块 ,然后进行块生长 ,找出其余的内部块和边界块 ,从而获得整个细胞核区域 .基于特征块的细胞核提取方法可以适应多种情况 ,能够快速、准确和完整地提取细胞核 .

It is a prerequisite to extract nuclear regions from blood microscopic images for quantitatively analyzing morphology of blood karyocytes. Nuclear color is nonhomogeneous. Global regional features are emphasized in human’s perception. Following this principle, authors propose a novel approach to extract nuclei based on feature blocks. A feature block for a nucleus is first identified in terms of principle color information of nuclei. Internal blocks and border blocks for a nucleus are detected based on the feature block. The feature block, internal blocks and border blocks belonging to a nucleus are used to extract an entire nucleus. Many applications show that the approach is effective and robust, especially for nonhomogeneous nuclear regions contained in blood microscopic images.