Saliency detection across spatial and frequency domains

Existing saliency detection algorithms work merely in one domain, i.e., spatial domain or frequency domain. This is not sufficient to process different types of images. To overcome this difficulty, we propose an algorithm that could deal with any type of images. Firstly, an image classifier is trained to classify different categories of images. Then, we adopt a strategy to process different types of images with different kinds of saliency detectors. Finally, an enhanced segmentation algorithm is employed to improve the quality of saliency map. Extensive simulation results show that our proposed method outperforms existing seven state-of-the-art algorithms.