Saliency detection improved by Principle Component Analysis and boundary scoring approach

Salient region detection is useful for many image processing applications, such as adaptive compression, object recognition, image retrieval, filter design, and image retargeting. In this paper, we propose a novel method to determine salient regions in images. Principle Component Analysis (PCA) is served as preprocessing for dimensionality reduction. It can reduce computational complexity and attenuate noise and translation error. Then, the local-global contrast is used to calculate distinctiveness. Finally, we take advantage of image segmentation to achieve full resolution saliency maps. Our proposed method is compared with the state-of-art saliency detection methods and yields higher precision and better recall rate.