Research And Implement Of Cell Image Segmentation Based On Edge Detect

The application of Image processing in the domain of medicine is concerned widely. Image processing is a valuable tool of clinical diagnosis and medical research by its applications include image segmentation, image identification and quantitative analysis. Image segmentation, as a key procedure of Image processing and the focus of researching of medical image processing, has been highly attention since 1970s.The segmentation of cell image and accurate extraction of cells' figure are the foundation of morphological analysis and quantitative analysis in the next approach.The paper demonstrates the definition and objective of Image segmentation. Firstly, Image segmentation is a process which segment entire image into several no superposed and nonempty subregions. Each subregion connects with others within. And the interior of subregion has the same or similar features. Feature of an Image means a remarkable attribute of an image. These features can be divided into two kinds, one kind is statistical feature, and another is visual feature. In another word, Image segmentation separates an image into several significant parts or regions by these two kinds of features. The paper describes ways of Image segmentation. In this paper, we sorts four kinds of ways of Image segmentation. First, segmentation based on threshold. If target pixels' gray level has obvious differences with background pixels', threshold segmentation would be an effective way to separate target from background. Secondly, segmentation based on edge detect. The basic feature of an Image is edge. It reflects the partial features' discontinuity. The fundamental warranties of edge point judgment are the first derivative's extremum and the second derivative's cross point of Image pixels. Thirdly, segmentation based on seeded region growing. Virtually, the seeded region growing segmentation aims to form a segmentation area by connect regions with same feature. Finally, segmentation based on using special tools. Tools could be used conclude Mathematical Morphology, fuzzy theory, neural networks and wavelet. According to the survey to the four kinds of ways of Image segmentation and image algorithm experiment in MATLAB, we summarized the relative merits of these traditional image processing algorithms. According to the features of cell image, the paper describe the way of cells' edge accurate detect. In this way, the paper use a new segmentation algorithm based on multiple partial thresholds presented by this paper. The segmentation algorithm is better on detail division and segmentation, and is well used when the image with multiple targets and complex background. Finally, the paperprovide experimental result of the way of cells' edge accurate detect. Compare with the traditional way, the result has a more accurate edge of cells' figure and maintains karyons' figure as much as possible.