A Technique Research On Automatically Classify RBC&WBC Image That Capable Of Touching-Cells

As the rapidly development of the IT, the requirements of the computer from human have greatly changed, the powerful arithmetic capability and the gigantic storage capacity can't meet people's desire. We hope to see that the computer can imitate some human activities especially in the mental field. Then the application boundary of computer have margined out times without number. The image processing, analysis and automated identification on Medical Image by computer play important role in the clinical diagnosis. The cells'amount, themselves' shape and the ratio of the WBC&RBC are one of the most based diagnoses in the clinical diagnosis by the analysis of Medical Image. But, this based diagnoses almost carry out manually, it is both waste time and energy. To solve this, we make a combination of the image processing technology, the pattern recognition technology and the data mining theory. Based on this we design a new algorithm which can automatically recognize the WBC&RBC and figure out the amount, ratio by input a microscopy cell images.In this paper, we focus on two problems. One is the automatic classifying the WBC, RBC and Touching Cells based on image segment. Another is the identifying and segmentation the touching cells.The main work and innovations of this paper are as follows:(1) For the first problem mentioned in previous, we quantize the feature vectors based on appropriate feature selection. To certify the rationality of the feature selection, we make some analysis base on the well known criterion on feature selection—the reliability, the independence and the distinguishability to testify.(2) The solution to the second problem is the innovation of this paper. We design a new algorithm that integrates the Hough-Cycle Detect and the Cluster Analysis together to carry out the identification and segmentation of touching-cells ,of which is enlightened by the geometric shape of the cells.