Study On Image Region Growing Algorithm’s Stability

Nowadays, the application of image processing technology has permeated into every aspects of human life. Image segmentation is one of the important parts in the image engineering, with the function of providing the basis for the follow-up work, such as image structure description, image recognition and image understanding. The main purpose of image segmentation is to segment an image into some regions which have certain relevant features.This paper first conducts a study on some popular image segmentation methods and discusses as well as analyzes the principles and features of typical image segmentation algorithms. And then makes an introduction about Gaussian Mixture Model (GMM). If we regard every region of image histogram as a single Gaussian Model, the image histogram will be considered as the result of repeated addition by many Single Gaussian Models. This paper will use Gaussian Mixed Model to fit image histogram and EM algorithm to estimate Gaussian parameters, thus to segment an image into some regions. In addition, this paper adds a constraint condition on the definition of traditional region to build a simple model of region based on Gaussian distribution. The expansion and improvement on the definition of region is the precondition for the stability of the result of region growing algorithm, which can reduce the randomness of region growing algorithm. From the definition of region, this paper puts forward stable region growing algorithm and automatic region growing algorithm. Finally, the experiments in this paper show that the selection of seed points and threshold has smaller effect on the result of segmentation algorithm. Thus, the segmentation algorithm can extract the stable region and have better robustness and need lesser manual interaction. A stable Region growing algorithm will have broad application in many fields such as pathological image segmentation, industrial image segmentation, extraction of the confidential information in safety image processing and satellite image segmentation.