低对比度图像目标分割的研究

低对比度图像目标分割的研究 Target Segmentation of Low Constrast Image 王思贤 WANG Si- xian 1 first-author 刘宗义 LIU Zong- yi 1 梅建新 MEI Jian- xin 1 武汉大学电子信息学学院,湖北,武汉,430072 武汉大学电子信息学学院,湖北,武汉,430072 College of Electronic Inform ation,Wuhan U niversity,Wuhan 0 72 College of Electronic Inform ation,Wuhan U niversity,Wuhan 0 72 利用图像分割技术 ,采用矩不变自动门限方法 ,对低对比度的精子显微图像进行了自动分割 ,实验表明 ,该方法能够在较短的时间内分割出目标和背景 ,同时能克服二值化过程中出现的断线和粘连等问题 ,且方法简单 ,容易实现 ,具有较好的实用价值 . Image segmentation is an important skill in image processing.In this paper the automatical segmention of multilple objects （ sperm cells） is dicussed.Because of poor quality of the image,some traditional automatical segmentations based on histogram can not exactly extract features of an object. Aimed at the inherent defection in moment invariant method,an improved method is proposed.This improved method can take into accountthe gray levels of the whole image and neighborhand feature of the pixels simultaneously,and it is easy to exatract features from an image.Compared with other method, firstly this method pocesses some advantages in the aspects of precission and speed for image segmentation and is available forthe automatic thresholding .Secondly,this method can getrid of adhension and rupture in the image segmentation.The method is especially useful for low contrast image,which is characterized by its good partition peculiarity,low computing cost,strong robust.In our experiments this method was used in the target segmentation of sperm microscopic image and satisfied results have been gotten.It can be used in other low contrastimages too. 显微图像 图像分割 阈值分割 对比度 microscopic image image segmentation threshold segmentation contrast 10.14188/j.1671-8836.2000.05.029 TP391.41 2000-05-01 2021-04-01 5