Study On The Segmentation Method Of Adhesion Cells Based On EmguCv

At present,computer technology is used by more and more people to analyze images of cell and lesion to help diagnose various diseases in medical field,such as the early detection of cancer.In the early stage,it is often difficult to detect because of no obvious physical signs,with large number of cancer cells that do not manifest themselves.However,the chance of cell lesion can be obtained by collecting the number of cells and parameters of cells in the image analysis of the human body,which can improve the detection probability of early cancer.Therefore,high magnification microscopic image analysis of cells plays an important role in the diagnosis of disease.We often see two or more cells that are very close or even overlapping in cell images.This phenomenon usually has a bad effect on the statistical analysis of cell parameters.Therefore,it is a very important step to divide adhesion cell image into a single cell images in the cell microimage processing,and also very important for the effectiveness of subsequent analysis and statistics.A variety of image segmentation algorithms in the cell microscopic image processing are proposed,such as gray threshold method,region growth method,edge detection method,etc.These algorithms all require the image to have one or more kinds of particular characteristics.There are some limitations in the actual image processing.No one kind of algorithm can perfectly segment any image.For example,the edge detection method divides the image based on the feature that the gray value at the boundary of the target image tends to change sharply,and the target object is separated from the image by the edge detection algorithm.Due to the existence of principle,which the edge detection and segmentation is based on the nature of different features between different regions in the image(such as the grayscale discontinuity of the region edges),the boundary between different regions is detected,so the obtained target region often appears.Dividing lines are discontinuous,even the result of wrong dividing lines.As is shown above algorithm,a EmguCv based segmentation method for adherent cells isproposed for microscopic images of human exfoliated cells.The algorithm first preprocessed to reduce the noise of the cell image.Then the cell image was first segmented by a simple watershed algorithm.And then the watershed algorithm is calculated to find the cell boundary,the gradient value of the cell image obtained to divide the image,and then the adhesion was fine by the method of expansion and corrosion of the cell boundary.The cell is divided.The experimental results show that the algorithm can successfully divide the two cells to adhesion together,and to a certain extent,if the cell cluster of multiple cells is not overlapping,it can also be segmented to a certain extent.