Models of Bottom-up Attention and Saliency

doi:10.1016/B978-012375731-9/50098-7

Neurobiology of Attention

2005, Pages 576-582

https://doi.org/10.1016/B978-012375731-9/50098-7 Get rights and content

ABSTRACT

Visually conspicuous, or so-called salient, stimuli often have the capability of attracting focal visual attention toward their locations. Several computational architectures subserving this bottom-up, stimulus-driven, spatiotemporal deployment of attention are reviewed in this chapter. The resulting computational models have applications not only to the prediction of visual search psychophysics, but also, in the domain of machine vision, to the rapid selection of regions of interest in complex, cluttered visual environments. We describe an unusal such application, to the objective evaluation of advertising designs.

References (0)

Cited by (129)

Structural deviations drive an uncanny valley of physical places
2022, Journal of Environmental Psychology
Certain built environments can decrease aesthetic appeal. For humans and objects, deviation from typical appearances leads to nonlinear appraisal characterized as the uncanny valley. The first time, it was explored whether an uncanny valley can be found for built environments. In Experiment 1, a cubic N-shaped function of uncanniness plotted against realism of built environments was found, indicating an uncanny valley. Quantitative and qualitative data indicate an association between uncanniness and structural anomalies. Experiment 2 explored distortions leading to uncanniness of indoor places. In Experiment 3, human presence decreased uncanniness of distorted indoor public places but increased uncanniness of private rooms. Taken together, the evidence indicates that deviations from familiar configural patterns drive uncanniness of built physical places. Thus, strong deviations from a built environment's predictable pattern decreases its aesthetic appeal.
Understanding the differential impact of children's TV on executive functions: a narrative-processing analysis
2022, Infant Behavior and Development
Citation Excerpt :
One such visual feature is flicker (i.e., luminance difference over time). Capturing attention in a bottom-up fashion, flicker appears to be one of the strongest exogenous drivers of attention in a scene, with its independent contribution to gaze location higher than the cumulative contribution of all features in models of visual saliency (Itti, 2005; Mital, Smith, Hill, & Henderson, 2011). A further stimulus level property which may constrain subsequent processing is Edge Density.
Evidence from multiple empirical studies suggests children’s Executive Functions are depleted immediately after viewing some types of TV content but not others. Correlational evidence suggests any such effects may be most problematic during the pre-school years. To establish whether “screen-time” is developmentally appropriate at this age we believe a nuanced approach must be taken to the analysis of individual pieces of media and their potential demands on viewer cognition. To this end we apply a cognitive theory of visual narrative processing, the Scene Perception and Event Comprehension Theory (SPECT; Loschky, Larson, Smith, & Magliano, 2020) to the analysis of TV shows previously used to investigate short-term effects of TV viewing. A theoretical formalisation of individual content properties, together with a quantitative content-based analysis of previously used children’s content (Lillard & Peterson, 2011; Lillard et al., 2015b) is presented. This analysis found a pattern of greater stimulus saliency, increased situational change and a greater combined presence of cognitively demanding features for videos previously shown to reduce children’s EF after viewing. Limitations of this pilot application of SPECT are presented and proposals for future empirical investigations of the psychological mechanisms activated by specific TV viewing content are considered.
ERP signatures of auditory awareness in cross-modal distractor-induced deafness
2021, Consciousness and Cognition
Previous research showed that dual-task processes such as the attentional blink are not always transferable from unimodal to cross-modal settings. This study investigated whether such a transfer can be stated for a distractor-induced impairment of target detection established in vision (distractor-induced blindness, DIB) and recently observed in the auditory modality (distractor-induced deafness, DID). A cross-modal DID effect was confirmed: The detection of an auditory target indicated by a visual cue was impaired if multiple auditory distractors preceded the target. Event-related potentials (ERPs) were used to identify psychophysiological correlates of target detection. A frontal negativity about 200 ms succeeded by a sustained, widespread negativity was associated with auditory target awareness. In contrast to unimodal findings, P3 amplitude was not enhanced for hits. The results support the notion that early frontal attentional processes are linked to auditory awareness, whereas the P3 does not seem to be a reliable indicator of target access.
The effect of load on spatial attention depends on preview: Evidence from a reading study
2018, Vision Research
Citation Excerpt :
In addition to models of eye movement control in reading, other models of eye movement control in tasks like scene processing or visual search can benefit from including effects of load on the focus of attention. To build their visibility map, these models normally use a visual field that is constrained by visual acuity but not the availability of spatial attention during the course of the fixation (e.g., Ghahghaei & Verghese, 2015; Itti, Rees, & Tsotsos, 2005; Najemnik & Geisler, 2005; Renninger, Verghese, & Coughlan, 2007). These models will benefit from considering the availability of spatial attention – as it depends on the time elapsed since the last saccade in addition to when the upcoming saccade is made- and the ongoing processing load.
The spatio-temporal distribution of covert attention has usually been studied under unfamiliar tasks with static viewing. It is important to extend this work to familiar tasks such as reading where sequential eye movements are made. Our previous work with reading showed that covert spatial attention around the gaze location is affected by the fixated word frequency, or the processing load exerted by the word, as early as 40 ms into the fixation. Here, we hypothesised that this early effect of frequency is only possible when the word is previewed and thus pre-processed before being fixated. We tested this hypothesis by preventing preview. We investigated the dynamics of spatial attention around the gaze location while the observer read strings of random words. The words were either always exposed (normal preview) or only exposed while being fixated (masked preview). We probed spatial attention when a target word with either high or low printed frequency – or low or high load, respectively – was fixated. The results confirmed that, early in a fixation, allocation of spatial attention 6 characters from the gaze was affected by the word’s frequency but only when the word was exposed before being fixated, so that processing of the word could start before it was fixated. Our results indicate that the ongoing processing load of a word is modulated by its pre-processing and affects the dynamics of covert spatial attention around the word once it is fixated.
Web page visual hierarchy: Examining Faraday's guidelines for entry points
2018, Computers in Human Behavior
Designers want to avoid creating “flat” visual experiences that do not communicate importance within an interface. Faraday's visual hierarchy model is often employed to predict the deployment of attention within web page interfaces (i.e., provide an importance ranking). However, only limited empirical evidence is available to validate its use. Across two studies, the model's predictive ability to determine entry points, as determined by the initial search phase, was examined. In Study 1, simple web pages were artificially manipulated to offer entry points as predicted by the model. In Study 2, current web pages containing greater complexity were coded for entry points. Across both studies, fixations were recorded while participants viewed the web pages. It was discovered that the model does a poor job predicting entry points. Moreover, it demonstrated a lack of sophistication by mainly selecting entry points on element size alone. It was found that both spatial position and distinctiveness can predict earlier attention engagement. Therefore, any updated visual hierarchy model ought to consider user expectations and visual salience. Critically, designers ought to stop employing this visual hierarchy model as its predictions can be misleading.
Consumer Neuroscience
2023, Consumer Neuroscience

View all citing articles on Scopus

View full text

CHAPTER 94 - Models of Bottom-up Attention and Saliency

ABSTRACT