边缘分类的椭圆检测在白细胞图像中的研究

Research on ellipse detection based on edge classificationin white blood cell image

针对血细胞图像中,白细胞形态各异,难以确定大小和位置信息的问题,提出一种基于边缘分类的椭圆检测方法。首先,在原白细胞图像的基础上使用K-means++彩色聚类算法进行分割预处理,有效消除了背景中红细胞与血小板的干扰。其次,通过边缘分类的椭圆检测算法处理分割后的白细胞图像,以获得白细胞的外轮廓信息。最后,提取得到白细胞的位置和大小信息参数。实验表明:边缘分类的椭圆检测与其他方法相比,其综合性能指标F值最高,在检测椭圆数目为3时,其综合性能指标F值与准确率P高达0.85,可以成功检测BCCD数据集中全部的8类不同形态的白细胞。

Aiming at the problem that the size and position of white blood cells in blood cell images are difficult to be determined due to their various forms,an ellipse detection method based on edge classification is proposed.Firstly,the K-means++color clustering algorithm is used for segmentation and pretreatment on the basis of the proleukocyte image,which effectively eliminates the interference between red blood cells and platelets in the background.Secondly,ellipse detection algorithm of edge classification is used to process the segmented white blood cell image to obtain the contour information of white blood cell.Finally,the location and size information parameters of white blood cells are extracted.The experimental results show that compared with other methods,edge classification ellipse detection method has the highest comprehensive performance index F value,and when the number of ellipse detection N=3,its comprehensive performance index F value and accuracy P are as high as 0.85,which can successfully detect all 8 types of different forms of white blood cells in BCCD dataset.