Image segmentation using transition region and K-means clustering

Several methods based on the transition region have been developed in image segmentation. Most are reported as effective methods with some limitations. Some are difficult to reduce background appearance, while others are difficult to find intact objects. A method which is capable of extracting a more intact transition region while reducing the background appearance is needed. In this study, we propose a novel method to extract a more intact transition region by combining the transition region with k-means clustering. First, a grayscale image is simplified to become a simple image by using k-means clustering. The simple image is a version of the original grayscale image with a lower gray level variation. Two transition regions are extracted: from the original image and the simple image. Then, the edge linking process is conducted based on the Quasi-Euclidean distance on the original image transition region by using the simple image transition region as a reference. Finally, region filling is done to get areas that are considered as objects. The proposed method is compared with two other transition region-based methods, and the experimental results show that the proposed method has the best performance in terms of image segmentation in general and the appearance of objects in the segmentation result.