低对比度图像目标分割的研究

Target Segmentation of Low Constrast Image

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.