Region-Based Image Segmentation Based on K-means and Probability Relaxation

In order to reduce over-segmentation,can take advantages of K-means which is simple,fast and able to deal with large database and probability relaxation.The method of combining K-means and probability relaxation is used in this paper.First apply K-means clustering method to segment the image pixels into different regiments.Then an iterative probability relaxation operation is applied in order to optimize the coarse segmentation to further segment the uncertain pixels according to their statistic properties.Experimental results indicate that proposed method is effective for image segmentation and object extraction.