Research On Segmentation Algorithms Of Blood Cell Microscopic Image

Blood smear microscopic examination is the basic and widely used examination method of blood cytology.Traditional white blood count and erythrocyte morphological analysis is based on manual smear and examined under a microscope by a qualified doctor, this is a very time consuming process and the test results depends on the observer’s expertise and experience, which make deviation be bigger.With image processing and target recognition technology is widely used in the field of biomedical,microscope inspection also gradually transition from manual method to automatic analysis.In recent years,the automatic detection system has become a new research hot spot in biomedical field,using the computer of high performance significantly improved the image processing speed,thus greatly shortens the clinical pathologic time of judgment.Microscopic examination of the automatic system in general can be divided into image acquisition,image processing and recognition process.The paper makes research on segmentation method of blood cell microscopic image.Considering the complicated blood cell microscopy image分割background,a lot of interference,cell adhesion and cell forms and so on,on the basis of image acquisition platform inherent mechanism,blood cell division of the paper can be divided into three steps:WBC(white blood cell)detection,WBC nuclear plasma separation and RBC(red blood cell)division.The main research work and achievements include:1 、 In WBC detection steps, making research on several commonly classical threshold segmentation method, introducing the principle of method and analyzing and comparing the advantages and disadvantages of detection methods respectively.Finally, this paper introduces the internal and external average contrast method and applied it to detect the WBC. Compared with other algorithm, it has stronger ability of tracking and detection and can get the ROI which contains WBC.2 、In WBC nuclear plasma separation steps: making research on the different color space of white blood cells and the different characteristics of images. This paper uses the OTSU and Region growing method to separate the WBC region of nucleus and cytoplasm. Using the information fusion theory, This paper proposes a kind of segmentation method of color white blood cell image based on HSI space. Theexperimental results show that this method can separate leukocyte nuclear plasma area accurately and can meet the requirements of medical image processing.3、In the RBC segmentation steps: making research on the basic principles of watershed transform and improved method. According to the characteristics of red blood cell image inherent combining advantages of the region and the gradient watershed transform, this paper put forward a kind of combination with distance transformation and edge gradient watershed segmentation algorithm. The experimental results show that this algorithm gets better segmentation results, faster real-time and better universal that could satisfy the requirement of medical image processing.