摘要
提出了一种变图像分辨率的染色血液白细胞目标图像提取方法.首先对采集到的染色血液细胞图像进行分辨率抽样并将图像从RGB表示变换为HS I表示;然后使用白细胞色饱和度阈值滤波和形态学滤波得到白细胞数目和位置;最后在白细胞位置按原始图像分辨率进行局部图像提取并使用神经网络方法进行图像分割得到目标图像.该方法充分保留了血液细胞图像本身固有的大量有用信息,提高了目标图像分割的完整性并减少了计算量,适用于对时间要求较高的复杂背景下血液图像的自动分析.
A new methodology that multi-resolution image segmentation method for while bolld cell is proposed in this paper. First, the captured image was sampled in low resolution and the RGB value of the image was transformed to HSI value. The threshold method and morphology method was used for segmentation to find the position white blood cell located, and then the local area was extracted as a subimage around the position mentioned above in original high rresolution and this subimage area was recognized by the method of artificial neural network. This method was fast, effective and can be used for automatic analysis of white blood cell in the case of time limited and the image with other complex objects as background. It has more benefits such as less computing steps and more precise in image recognition than the normal method in one resolution.
出处
《南开大学学报（自然科学版）》
CAS
CSCD
北大核心
2006年第6期59-63,共5页
Acta Scientiarum Naturalium Universitatis Nankaiensis
基金
天津市科委自然科学基金(023615211)
天津市教委科学基金(20010204)
关键词
彩色图像识别
图像分割
细胞图像识别
人工神经网络
color image recognition
image segmentation
cell image recognition
artifical nueural network
