【正文】 to some people the characteristics of definition, through the transform to get, such as image histogram, moments, spectrum, etc.。 Image visual characteristics is refers to person visual sense can be directly by the natural features, such as the brightness of the area, and texture or outline, etc. The two kinds of characteristics of the image into a series of meaningful goal or regional process called image segmentation.The image is the basic characteristics of edge, the edge is to show its pixel grayscale around a step change order or roof of the collection of those changes pixels. It exists in target and background, goals and objectives, regional and region, the yuan and the yuan between, therefore, it is the image segmentation dependent on the most important characteristic that the texture characteristics of important information sources and shape characteristics of the foundation, and the image of the texture characteristics and the extraction of shape often dependent on image segmentation. Image edge extraction is also the basis of image matching, because it is the sign of position, the change of the original is not sensitive, and can be used for matching the feature points.The edge of the image is reflected by gray not continuity. Classic edge extraction method is investigation of each pixel image in an area of the gray change, use edge first or second order nearby directional derivative change rule, with simple method of edge detection, this method called edge detection method of local operators.The type of edge can be divided into two types: (1) step representation sexual edge, it on both sides of the pixel gray value varies significantly different。 (2) the roof edges, it is located in gray value from the change of increased to reduce the turning point. For order jump sexual edge, second order directional derivative in edge is zero cross。 For the roof edges, second order directional derivative in edge take extreme value.If a pixel fell in the image a certain object boundary, then its field will bee a gray level with the change. The most useful to change two features is the rate of change and the gray direction, they are in the range of the gradient vector and the direction to said. Edge detection operator check every pixel grayscale rate fields and evaluation, and also include to determine the directions of the most use based on directional derivative deconvolution method for masking.Digital image processing technique has been widely applied to the biomedical field, the use of puter image processing and analysis, and plete detection and recognition of cancer cells can help doctors make a diagnosis of tumor cancers. Need to be made in the identification of cancer cells, the quantitative results, the human eye is difficult to accurately plete such work, and the use of puter image processing to plete the analysis and identification of the microscopic images have made great progress. In recent years, domestic and foreign medical images of cancer cells testing to identify the researchers put forward a lot of theory and method for the diagnosis of cancer cells has very important meaning and practical value. Cell edge detection is the cell area of ??the number of roundness and color, shape and chromaticity calculation and the basis of the analysis their test results directly affect the analysis and diagnosis of the disease. Classical edge detection operators such as Sobel operator, Laplacian operator, each pixel neighborhood of the image gray scale changes to detect the edge. Although these operators is simple, fast, but there are sensitive to noise, get isolated or in short sections of a continuous edge pixels, overlapping the adjacent cell edge defects, while the optimal threshold segmentation and contour extraction method of bining edge detection, obtained by the iterative algorithm for the optimal threshold for image segmentation, contour extraction algorithm, digging inside the cell pixels, the last remaining part of the image is the edge of the cell, change the processing order of the traditional edge detection algorithm, by MATLAB programming, the experimental results that can effectively suppress the noise impact at the same time be able to objectively and correctly select the edge detection threshold, precision cell edge detection.2. Edge detection of MATLABMATLAB image processing toolkit defines the edge () function is used to test the edge of gray image.(1) BW = edge (I, method), returns and I size binary image BW, including elements of 1 said is on the edge of the point, 0 means the edge points. Method for the following a string of:1) soble: the default value, with derivative Sobel edge detection approximate measure, to return to a maximum gradient edge。2) prewitt: with the derivative prewitt approximate edge detection, a maximum gradient to return to edge。3) Roberts: with the derivative Roberts approximate edge detection margins, return to a maximum gradient edge。4) the log: use the Laplace operation gaussian filter to I carry filtering, through the looking for 0 intersecting detection of edge。5) zerocross: use the filter to designated I filter, looking for 0 intersecting detection of edge.(2) BW = edge (I, method, thresh) with thresh designated sensitivity threshold value, rather than the edge of all not thresh are ignored.(3) BW = edge (I, method thresh, direction, for soble and prewitt method specified direction, direction for string, including horizontal level said direction。 Vertical said to hang straight party。 Both said the two directions (the default).(4) BW = edge (I, 39。log39。, thresh, log sigma), with sigma specified standard deviation.(5) [BW, thresh] = edge (...), the return value of a function in fact have multiple ( BW and thresh ), but because the brace up with u said as a matrix, and so can be thought a return only parameters, which also shows the introduction of the concept of matrix MATLAB unity and superiority.3. Last wordMATLAB has strong imag