Research On Segmentation Algorithms Of Wall-pasted Cells Image

Precise counting of living cells is necessary and important in vitro anti-virus experiment.The counting accuracy of living cells is restricted by the traditional methods of examining theantivirotic medicine criteria. Image processing technology is a new powerful technology forcell counting and analysis and is able to improve facticity of medicine screening experimentresult. In this paper we use three image segmentation methods to segment the cell image andcompare their advantages and disadvantages:(1) Region growing method. We first use region growing method to segment cell imageand then apply the morphology close operation to the segmented image. The granulesin the inner of the cell are cleared and the rupture of the boundary is connected.Experiments show region growing method is fast and accurate for cell imagesegmentation.(2) Parametric active contour (PAC) model. We introduce several PAC models(Kass model,Baloon model, GVF model) and discuss their advantages and disadvantages. At last weprove that the PAC is not fit for the segmentation of the whole cell image byexperiments.(3) We use a geometric active contour model which is based on Mumford function andLevel set to segment cell image and discuss its advantage and disadvantage. We useGabor filter to enhance the weak boundary and to weaken the inner of the cell indifferent angle. We fuse the enhanced images to one. At last we use the geometricactive contour model to the enhanced image and get a good segmenting result.We found region method is fast accurate for cell image segmentation. Parametric activecontour is not fit for the whole image, but it is fit for segmentation of single cell in the image.Geometric active contour is antinoise and fit for the cell image segmentation.