Research On Segmentation For Overlapped Circular Cell Images

With the rapid development of computer technology, computer image processing technology has rapidly penetrated into all walks of human life. Cell inspection and analysis is of importance to human, especially in the aspect of biological cell medicine. However, due to the equipment error and uniform-illuminated and so on, sampled image may be blurred and noise polluted, as well as visual fatigue caused by long-term observation with the naked eye may affect the results of analysis and judgments.In this paper, use digital image processing techniques to detect and analyse the biological cell images, achieved automation of this process, greatly reduced the consumption of human and improved the accuracy of the results. Meanwhile, develop an adaptive segmentation software for cell images. The main works carried on are as follows:1. Comprehensively overview the related research and development of domestic and international for image segmentation and cell adhension, analyse and compare the common image segmentation methods and adhension methods, summer up the advantages and disadvantages and problems of these methods.2. Propose an adaptive segmentation method for cell image. This method combined with features of biological cell image, use multi-color and single-channel selection to determine an corresponding segmentation method, realize adaptive segmention with strong universality and robustness.3. Research and programs to achieve the edge tracking algorithm based on eight-neighborhood chain coding, overlook complex texture inside the biological cells, lay a good foundation for circle detection used Hough transform based on gradient of edge.4. For the situation of cells pile up and adhension severe, a cell adhension segmention method used circle detection is proposed based on gradient information, for the most fungi cells are circular take the edge profile tracked to carry on circle detection, remove false cells caused by Hough transform determined by established authenticity criterion of cells, achieve accurate segmentation.5. Use Visual C++ development tool to design an adaptive segmentation software for cell images, show the histogram of each channel for RGB and HIS color space, display the color gradient of OTSU algorithm, output functions of adaptive segmentation and so on, use Matlab Simulation to realize isolation and detection for cells heavily overlapping and adhension. The algorithm has universal adaptability and strong robustness.