Image segmentation algorithm based on hierarchal granulation model of variable precision

In order to deal with space correlation in character of image information, a new image segmentation algorithm is proposed which is based on variable precision hierarchy granular model. Firstly, we introduce classifying error precision into knowledge granulation, and construct granular structure of the image on various levels of confidence and quality of classification. Next, on the basis of the requirement of segmentation precision, we choose unit granular layer and analyze the importance of different grey levels on it further. Finally, the equivalent relations defined by the dissimilarities are used to implement the combination of the similar regions and the image segmentation is accomplished. The algorithm is applied to image segmentation tests. The experimental results indicate that it not only improve the parallel computation of the image and reduce complexity of space and time, but also provide new thoughts of knowledge granulation in image process.