GRAPH-BASED IMAGE SEGMENTATION WITH BAG-OF-PIXELS

Image segmentation can help us with a better understanding of the image content, and thus plays an important role in many fields of image processing including image retrieval, object recognition and image compression. We propose a novel method based on an efficient Graph-Based segmentation after analyzing its irrationalities. The proposed method with bag-of-pixels is simple and effective. In the evaluation using a large public image database that contains a lot of natural images, our method outperforms the efficient graph-based segmentation method, yields better segmentation results and shortens the running time.