Feature combination strategies for saliency-based visual attention systems

By Laurent Itti, Christof Koch. J. Electronic Imaging, 2001

484 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Computer Vision and Pattern Recognition, Human-Computer Interaction

View Paper: https://www.semanticscholar.org/paper/1a535dddfdab579c8f0556ad6ce04b5089f78a96

How researchers see this paper

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

In Cognitive Neuroscience

1. Laurent Itti
2. Vidhya Navalpakkam
3. Simone Frintrop
4. Ali Borji
5. Erich Rome
6. Henrik Iskov Christensen
7. Pietro Perona
8. O. Le Meur

In Computer Vision and Pattern Recognition

1. Shuicheng Yan
2. Luc Van Gool
3. Antonio Torralba
4. Ming-Hsuan Yang
5. Ali Borji
6. Dacheng Tao
7. Xiaogang Wang
8. Yi Yang

In Human-Computer Interaction

1. Laurent Itti
2. Vidhya Navalpakkam
3. O. Le Meur
4. Simone Frintrop
5. Erich Rome
6. Henrik Iskov Christensen
7. Christof Koch
8. Shuicheng Yan

Citations in the corpus

• Computational modelling of visual attention (2001, background): “First, different features contribute with different strengths to perceptual saliency 53 ,and this relative feature weighting can be influenced according to the demands of the task through top-down modulation 38,54 and through training 55–58 .”
• Saliency Detection by Multitask Sparsity Pursuit (2012, background): “The prior knowledge on scene context or specific objects is also helpful for identifying saliency regions, leading to the exploration of combining bottom-up and top-down information [17], [22], [23], [32] for saliency detection.”
• Multi-scale Deep Learning Architectures for Person Re-identification (2017, background): “Deep saliency modelling Visual saliency has been studied extensively [19, 20].”
• Is bottom-up attention useful for object recognition? (2004, background): “N ´¡µ is an iterative normalization operator (for details see [13]).”
• Depth Matters: Influence of Depth Cues on Visual Saliency (2012, background): “Another important aspect brought out in Table 2 a multiplicative modulating effect explains the influence of depth on saliency better than a linear weighted summation model, the latter has been used popularly in literature to combine results saliency maps derived from individual features [28].”
• Predicting visual fixations on video based on low-level visual features. (2007, background): “The most recent is the work of L. Itti et al. (Itti & Koch, 2001) in which three bottom-up combination strategies are proposed: (1) the simple normalized summation (called NS in the following) is the simplest method.”
• Modeling the influence of task on attention. (2005, methodology): “map ( Itti & Koch, 2001b ) is expected to drive those”
• An Integrated Model of Top-Down and Bottom-Up Attention for Optimizing Detection Speed (2006, background): “This is an unsolved challenge as yet, since most models of top-down attention are sub-optimal heuristics and driven by knowledge of the desired target only [32, 19, 13, 10], while ignoring the contribution due to knowing the distracting background.”
• Computational visual attention systems and their cognitive foundations: A survey (2010, methodology): “One of the currently best-known attention systems is the Neuromorphic Vision Toolkit (NVT) (Figure 11), a derivative of the Koch–Ullman model, which is steadily kept up to date by the group around Itti [Itti et al. 1998; Itti and Koch 2001a; Navalpakkam and Itti 2006a].”
• Modeling attention to salient proto-objects (2006, unknown): “(7c) N (·) is an iterative, nonlinear normalization operator, simulating local competition between neighboring salient locations (Itti & Koch, 2001).”