Study On Segmentation Of Clustering Cell Image Based On Contour Tracing

The processing and analysis of cell image has a wide application in medical treatment, such as the colony, blood cell and etc. After these cells being segmented from the background, we can count the number, extract their features (eg. area, perimeter, diameter, color etc.), and then we can analyze.Segmenting is a key step in this process. The segmenting result has a direct influence on the accuracy of counting and following processing. However, segmenting is also a classic problem in image engineering. Though numerous domestic and international scholars have extensive lucubrated into this subject and have presented many effective algorithms, there is still no one algorithm that can be applied to any test image and its segmenting result is satisfied. Especially in cell images, the cells tend to cluster into each other and they are more difficult to be segmented. For this, based on the analysis of the presented segmenting algorithms, a new segmenting algorithm to improve the accuracy and speed is presented.Firstly, the paper introduces the foundation and study significance of the topic. After analysing the shortcoming of presented cell image segmenting algorithms, the study topic is putted forward. Then the cause of the shortcoming is studied in detail. At last, the segmenting algorithm of clustering cell image is presented.(1) Aiming at the problem that the clustering region whose axis direction is neither horizontal nor vertical can not be segmented by the presented algorithm early, an improved segmenting algorithm is presented in the paper. The rotating operation is done before the segmentation to make the direction of the axis of the clustering region to be horizontal or vertical.(2) Aiming at the problem that there are holes in the clustering region, the inner-contour tracing is presented. When the clustering has hole, its outer-contour and inner-contour are both traced in a pass of tracing operation for the segmentation algorithm based on contour tracing..(3) Aiming the problem of "over-segmented" and "under-segmented", the true2-D Euclidean distance is used to determine the number of kernel of cell in the segmenting algorithm, based on contour tracing, a segmenting algorithm by using of the Euclidean distance transformation is presented. When the number of kernel of the clustering block is more than one, then the algorithm does the segmenting, otherwise, the algorithm does not.The experiment shows the segmenting algorithm presented has improved the segmenting result effectly.