The Research Of Image Segmentation Method Based On Threshold Selection

Image segmentation is an important and typical problem of digital image processing. It is one of the most important and fundamental field in image processing and low level vision of computer vision field, and is the premises of image vision analysis and pattern recognition. At present, image segmentation technology has widely applied in many fields, such as industry automatization,online production test,produce process control,document image processing,remote sensing and biomedicine image analysis,bodyguard surveillance and military,sport,agriculture engineering and so on. Image segmentation based on the selections of image thresholding methods has characteristic of simple and good computational performance and nice stabilization, so it is the most basal segmentation technology and application widely. However, image segmentation results depend heavily on the chosen image thresholding methods.This paper focuses mainly on some characteristic of the researches of image segmentation algorithms based on the selections of image thresholding methods and its application. Our work includes the following subjects:1) After studying and analyzing the principle of Parzen-window nonparametric estimation method and the entropy-based thresholding method, we incorporate the Parzen-window method and the entropy principle, then the novel image thresholding method PWET is proposed. The experimental results also demonstrate the success of our method here through compared with the traditional KSW-entropy method.2) Based on the Parzen-window estimation method, the novel thresholding method PWT is proposed. The method utilizes the coordinate space information of the pixels of the image, and incorporates with the histogram of the image, estimates the space probability distribution density function of the image gray values with Parzen-window method, then the novel criterion function is designed. The experimental results demonstrate the success of our method here through compared with the traditional OTSU method,minimum error thresholding method (MET method for simplicity) and KSW-entropy method.