Exploiting local and global patch rarities for saliency detection

By Ali Borji, Laurent Itti. 2012 IEEE Conference on Computer Vision and Pattern Recognition, 2012

387 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Sensory Systems

View Paper: https://www.semanticscholar.org/paper/5e3d92c27e2fbb407243a6279dfafe9260a3732d

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

1. Ming-Ming Cheng
2. Ming-Hsuan Yang
3. Luc Van Gool
4. Ling Shao
5. Dacheng Tao
6. Xiaogang Wang
7. Yi Yang
8. Huchuan Lu

In Sensory Systems

1. Ming-Ming Cheng
2. Ali Borji
3. Ling Shao
4. Huaizu Jiang
5. Qi Zhao
6. Junwei Han
7. Jianbing Shen
8. Jia Li

Citations in the corpus

• Two-Stage Learning to Predict Human Eye Fixations via SDAEs (2016, methodology): “As a milestone work, Itti et al. [17] invented the center-surround difference operator to derive the contrasts in color, intensity, and orientation.”
• Visual Saliency Transformer (2021, background): “Among the three soft split steps, the patch sizes are set to k = [7, 3, 3], the overlappings are set to s = [3, 1, 1], and the padding sizes are set to p = [2, 1, 1].”
• A Simple Pooling-Based Design for Real-Time Salient Object Detection (2019, methodology): “Traditional methods [11, 25, 14, 31, 2, 12, 41, 3] mostly rely on hand-crafted features to capture local details and global context separately or simultaneously, but the lack of high-level semantic information restricts their ability to detect the integral salient objects in complex scenes.”
• Hierarchical Image Saliency Detection on Extended CSSD (2016, methodology): “The global method results in (c) also cannot clearly distinguish among regions.”
• Visual saliency estimation by nonlinearly integrating features using region covariances. (2013, result): “As compared to similar models that use local patch-based strategies (Borji & Itti, 2012; Duan, Wu, Miao, Qing, & Fu, 2011; Goferman et al., 2010; Seo & Milanfar, 2009), the main novelty of our model comes from using covariance descriptors of the regions to represent their visual characteristics.”
• Unsupervised Object Keypoint Learning using Local Spatial Predictability (2020, background): “An interesting distinction is between local and global saliency, where the former is concerned with (dis)similarity between local image regions and the latter focuses on how (in)frequently information content occurs across an entire dataset (Borji & Itti, 2012).”
• RGBD Salient Object Detection: A Benchmark and Algorithms (2014, background): “Formally, for any patch P , its feature vector is defined as f = [c, l, r,d] , where c is the average CIELab color value of pixels in P since CIELab can better approximate human color perception [5], l represents center location of pixels in P on the image plane, r is the region size defined by the ”
• SALICON: Reducing the Semantic Gap in Saliency Prediction by Adapting Deep Neural Networks (2015, methodology): “We also report the Human Inter-Observer (IO) score [1].”
• Saliency Detection: A Boolean Map Approach (2013, methodology): “(b) Saliency maps estimated by (from left to right) AIM [6], LG [3] and our method.”
• Salient object detection: A survey (2014, background): “patches, while their spatial distances are taken into account. Similarly, Borji and Itti computed local and global patch rarity in RGB and LAB color spaces and fused them to predict fixation locations [26]. In a recent work [149], Margolin et al. propose to define the uniqueness of a patch by measurin”
• Salient Object Detection: A Discriminative Regional Feature Integration Approach (2013, methodology): “Recent developments include using isocentric curvedness and color [44], adopting image histogram [32], quaternion-based spectral analysis [40], utilizing depth cues [30], multitask sparsity pursuit [29], statistically modeling [42], exploring patch rarities [6], combing bottom-up and top-down featur”
• What do eyes reveal about the mind?: Algorithmic inference of search targets from fixations (2015, background): “While the allocation of attention is often task-driven, it can also be guided by bottom-up and stimulusdriven cues [80,42,40,64,39,4,3,10].”
• Unsupervised Salience Learning for Person Re-identification (2013, background): “Different from general image salience detection methods [4], our salience is especially designed for human matching, and has the following properties.”
• How to Evaluate Foreground Maps (2014, methodology): “Common methods to integrate this into a single measure are Area-Under-the-Curve (AUC) [7, 12, 15] and Average-Precision (AP) [10, 23].”