Context-aware saliency detection

By Stas Goferman, Lihi Zelnik-Manor, Ayellet Tal. 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010

2,366 citations recorded by the source corpus.

We propose a new type of saliency – context-aware saliency – which aims at detecting the image regions that represent the scene. This definition differs from previous definitions whose goal is to either identify fixation points or detect the dominant object. In accordance with our saliency definition, we present a detection algorithm which is based on four principles observed in the psychological literature. The benefits of the proposed approach are evaluated in two applications where the context of the dominant objects is just as essential as the objects themselves. In image retargeting we demonstrate that using our saliency prevents distortions in the important regions. In summarization we show that our saliency helps to produce compact, appealing, and informative summaries.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/32e072aa603a64ca5b40fdc6b4c94c951166a310

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

New Contour Cue-Based Hybrid Sparse Learning for Salient Object Detection (2019, methodology): “…method (SARS) on four benchmark datasets along with 26 state-of-the-art saliency detection models, which for simplicity we denote as BMS [10], CA [52], CB [35], CGVS [53], CRF [15], DSR [20], FES [54], FT [16], GMR [30], GR [36], GS [28], HS [55], LC [56], LEGS [42], LMLC [41], LPS [57], MBS…”
Leveraging stereopsis for saliency analysis (2012, methodology): “(d) Saliency [11] (e) Saliency [14] (f) Saliency [1]”
Saliency in VR: How Do People Explore Virtual Environments? (2016, background): “Top-down models take higher-level knowledge of the scene into account such as context or specific tasks [16, 22, 25, 35, 55].”
We get the algorithms of our ground truths: Designing referential databases in digital image processing (2017, methodology): “Other proposed algorithms included Achanta et al. (2009); Chang et al. (2011); Goferman et al. (2012); Shen and Wu (2012); Wang et al. (2010).”
Efficient Salient Region Detection with Soft Image Abstraction (2013, methodology): “(a) original (b) IM[39] (c) SeR[45] (d) SUN[52] (e) SEG[43] (f) AIM[8] (g) SWD[20] (h) RC[16] (i) CA[25] (j) MZ[37]”
Depth saliency based on anisotropic center-surround difference (2014, methodology): “We choose three state-of-the-art methods work on color images namely CNTX [9], RC [10], PCA [11] and three depth saliency methods namely CURV [16], SS [13], DP [14] for comparison.”
Saliency Detection in the Compressed Domain for Adaptive Image Retargeting (2012, background): “In [16], Goferman et al. built a content-aware saliency detection with the consideration of the contrast from both local and global perspectives.”
Structure-Measure: A New Way to Evaluate Foreground Maps (2017, methodology): “nk the SM generated by a state-of-the-art model higher than a generic map. We count the number of times a generic map scored higher than the mean score generated by the ﬁve state-ofthe-art models (CA [19], CB [23], RC [13], PCA [35], SVO [9]). The mean score provides an indication of model robustnes”
Background Prior-Based Salient Object Detection via Deep Reconstruction Residual (2015, methodology): “It can be observed that, compared with PD, GBMR, GS-S, GS-G, BLSM, and CNTX, the proposed method can highlight the salient region more uniformly.”
A Stagewise Refinement Model for Detecting Salient Objects in Images (2017, neutral): “, [14, 9, 40, 24, 57, 28, 7].”
Large-Scale Optimization of Hierarchical Features for Saliency Prediction in Natural Images (2014, methodology): “To ro nt o Chance – 0.4988 – 4.1171 – 0.3040 – -0.0038 Cntr 0.7836 – 2.5828 – 0.4116 – 0.8180 – CovSal 0.8147 0.7616 2.3712 3.1187 0.5122 0.3812 1.4587 1.0090 ICL 0.8169 0.7807 2.4190 2.9943 0.5253 0.4333 1.6199 1.2483 Signature 0.8188 0.7903 2.3776 2.9054 0.5311 0.4658 1.5888 1.4464 GBVS 0.8274 0.8”
Visual Saliency Based on Scale-Space Analysis in the Frequency Domain (2013, neutral): “F”
Saliency Detection by Multitask Sparsity Pursuit (2012, methodology): “…models 1 for comparison, which are the Itti model (IT) [2], graph-based visual saliency (GBVS) [15], context aware-based saliency detection (CSD) [14], self-information (SI) [11], the sparse-coding-based method (SCSP) [30], saliency using natural statistics [46], and other two methods using…”
Saliency Optimization from Robust Background Detection (2014, neutral): “This is called contrast prior and is used in almost all saliency methods [25, 19, 22, 6, 16, 14, 17, 26, 28, 8, 9, 29, 3].”
Predicting the Driver's Focus of Attention: The DR(eye)VE Project (2017, positive): “This can be achieved by either merging saliency maps at different levels of scale and abstraction [23], or including a-priori cues about relevant objects for the task at hand [14], [21], [63].”
Salient object detection: A survey (2014, background): “e co-saliency by exploiting the additional repeatedness property across multiple images. Speciﬁcally, the cosaliency score of a pixel is deﬁned as the multiplication of its traditional saliency score [39] and its repeatedness likelihood over the input images. Fu et al. [121] propose a cluster-based ”
SuperCNN: A Superpixelwise Convolutional Neural Network for Salient Object Detection (2015, methodology): “The comparison is conducted between the proposed method and the context-aware saliency detection method CA (Goferman et al. 2010), using the non-homogeneous resizing technique proposed by Avidan and Shamir (2007).”
Salient Object Detection: A Benchmark (2015, neutral): “at may contain objects from any object category (a.k.a, category independent object proposals). # Model Pub Year Code Time(s) Cat. 1 AC [66] ICVS 2008 C .129 Detection 2 FT [18] CVPR 2009 C .072 3 CA [67] CVPR 2010 M + C 40.9 4 MSS [68] ICIP 2010 C .076 5 SEG [69] ECCV 2010 M + C 10.9 6 RC [70] CVPR”