Integration of the saliency-based seed extraction and random walks for image segmentation

In this paper, a novel automatic image segmentation method is proposed. To extract the foreground of the image automatically, we combine the region saliency based on entropy rate superpixel (RSBERS) with the affinity propagation clustering algorithm to get seeds in an unsupervised manner, and use random walks method to obtain the segmentation results. The RSBERS first applies entropy rate superpixel segmentation method to split the image into compact, homogeneous and similar-sized regions, and gets the saliency map by applying saliency estimation methods in each superpixel regions. Then, in each saliency region, we apply the affinity propagation clustering to extract the representative pixels and obtain the seeds. A relabeling strategy is presented to ensure the extracted seeds inside the expected object. Additionally, in order to enhance the effects of segmentation, a new feature descriptor is designed using the covariance matrices of coordinates, color and texture information. Experiments on publicly available data sets demonstrate the excellent segmentation performance of our proposed method.