Research And Implementation Of Image Segmentation Algorithm

Image segmentation is a critical image analysis technique,the purpose of which separates the object into regions with special character and extracts interesting region, the result of the image segmentation is the basis of the feature extraction and pattern Recognition.The research of the image segmentation is the hot spot and focus of the image technique,hundreds of algorithms have been presented in the literature, although those algorithms are to some extent successful,image segmentation is still far from solved,the research in this area still faces many challenges.The thesis describes ways of image segmentation,we sorts three kinds of ways of image segmentation:First,segmentation based on edge detection,the basic feature of an image is edge,the fundamental warranties of edge point judgment are the first derivative's extreme value and the second derivative's cross point of image pixeis. The thesis separates the cell images,compares the advantages and disadvantages of a variety of edge detection operator from the actual results;Secondly,segmentation based on region,if target pixels' gray level has obvious differences with background pixels',threshold segmentation would be an effective way to separate target from background;The seeded region growing segmentation aims to form a segmentation area by connect regions with same feature;Thirdly,segmentation based on using special tools,tools conclude wavelet,Mathematical Morphology,neural networks and fuzzy theory.The method of the wavelet combing with histogram is researched, making use of the multi-scale character of wavelet,we can choose different thresholds at different scale,this may make the threshold more reasonable, experimental results show that the method makes the segmentation have good effect. According to the survey to the three kinds of ways of image segmentation and image algorithm experiment,I summarized the relative merits of these traditional image segmentation algorithms and applicable environment.A modified partial adaptive segmentation algorithm was proposed to separate the cell images efficiently.First,it divided the whole image into several parts,then computed the grads histogram each parts and performed Otsu segmentation,at last,it considered the effect of yawp in the image and removed the yawp with morphologic yawp filter.The experimental results show that the algorithm can well differentiate and separate detail of the cell image,Compare with the traditional way,the result has a more accurate edge of cells' figure and maintains the greatest degree of nuclear outline of cells,and apply to the image that inconspicuous between background and objects and being some yawp.