针对胃上皮癌变细胞结构复杂和粘连现象严重的特点,提出了一种将多尺度小波变换和改进的大津法相结合的分割方法.该方法对图像进行多分辨率分解,通过小波分解的高频信息计算出独立峰的个数,利用改进的大津法确定相邻独立峰之间的最优阈值.实验结果表明,该方法具有计算简单、自适应性强、精确度高的优点,能有效地实现胃上皮肿瘤细胞图像的分割.%In view of characteristics of complex structure and serious adhesion phenomenon of Gastric epithelium cancerous cells, a new segmentation based on multi-scale wavelet analysis integrated with improved Otsu is brought forward. Firstly the method gets multi-resolution decomposition of images. Then using the high-frequency of wavelet decomposition the number of independent peaks are calculated. Finally, using improved Otsu the optimal threshold of the adjacent independent peaks are determined. Experiment indicts that this method is simple, the self-adaptability and effective. Segmentatioin's results is satisfied with demand of gastric epithelial cells' segmental demand.
展开▼