Saliency Optimization from Robust Background Detection

Abstract

Recent progresses in salient object detection have exploited the boundary prior, or background information, to assist other saliency cues such as contrast, achieving stateof-the-art results. However, their usage of boundary prior is very simple, fragile, and the integration with other cues is mostly heuristic. In this work, we present new methods to address these issues. First, we propose a robust background measure, called boundary connectivity. It characterizes the spatial layout of image regions with respect to image boundaries and is much more robust. It has an intuitive geometrical interpretation and presents unique benefits that are absent in previous saliency measures. Second, we propose a principled optimization framework to integrate multiple low level cues, including our background measure, to obtain clean and uniform saliency maps. Our formulation is intuitive, efficient and achieves state-of-the-art results on several benchmark datasets.

Correspondence Driven Saliency Transfer

“…fold ranking (MR13) [15], saliency model via robust background detection (wCtr14) [18], saliency model via cellular automata (BSCA15) [21], and saliency model via bootstrap 2 . .…”

“…We aim to simultaneously use foreground and background samples into a graph based segmentation method to obtain spatially consistent For the second principle, we measure the saliency of a region by its shortest distance to the boundary regions on geodesics. Geodesic distance is a powerful measurement for saliency detection [18], [31]. Additionally, our correspondence-based saliency offers an indication of a boundary region whether it belongs to the background or not.…”

“…In [15], the saliency of image regions was measured by their relevance to the image boundary via a manifold ranking scheme. The work in [18] constructed a robust boundary prior based on boundary connectivity. Qin et al [21] used the clustered boundary seeds into a cellular automata.…”

“…Wei et al [31] exploited 2 boundary prior by noting that the image boundary regions are more likely to belong to the background. Similarly, many follow-up studies [14]- [16], [18], [21] were proposed to treat image boundaries as background samples. To improve the performance, these approaches also explore more robust boundary priors.…”

“…Top-down approaches [7]- [10] are goal-directed and usually adopt supervised learning with a specific class. Most of the saliency detection methods are based on bottomup visual attention mechanisms [11]- [15], [17], [18], [21], which are independent of the knowledge of the content in the image and utilize various low level features, such as intensity, color and orientation. Those bottom-up saliency models are generally based on different mathematical formulations of center-surround contrast or treat the image boundary as the background.…”