摘要
为了从细胞核图像上鉴别癌细胞、良性肿瘤细胞、正常细胞,本文使用了统计方法实现了计算机细胞图像分类。计算机首先对输入图像进行两项预处理;核图像亮度归一化及核图像尺寸归一化,然后计算出核图像的两个特征值;平均颗粒面积及平均颗粒数目。这两个特征值能将70个结肠细胞及肺细胞的性质正确地区分开来。 本文重点介绍了核图像尺寸的归一化及两个特征值的简便算法。在核图像尺寸归一化及计算平均颗粒面积的过程中使用了形态滤波器。
This paper used statistics to do image classification for cell nuclei. Two kinds of normalizations: gray level normalization. nuclei image size normalization and two feature values' average granulation number average granulation erea were developed based on morphlogic filter. Measuring 70 cell samples showed that the two feature values can correctly discriminate three kinds of cell: malignate. binign. normal. Size normalization and two simple algorithms for finding out the two feature values will be described in this paper.
出处
《通信学报》
EI
CSCD
北大核心
1993年第3期73-77,共5页
Journal on Communications