Eye guidance in natural vision: Reinterpreting salience

Abstract

Models of gaze allocation in complex scenes are derived mainly from studies of static picture viewing. The dominant framework to emerge has been image salience, where properties of the stimulus play a crucial role in guiding the eyes. However, salience-based schemes are poor at accounting for many aspects of picture viewing and can fail dramatically in the context of natural task performance. These failures have led to the development of new models of gaze allocation in scene viewing that address a number of these issues. However, models based on the picture-viewing paradigm are unlikely to generalize to a broader range of experimental contexts, because the stimulus context is limited, and the dynamic, task-driven nature of vision is not represented. We argue that there is a need to move away from this class of model and find the principles that govern gaze allocation in a broader range of settings. We outline the major limitations of salience-based selection schemes and highlight what we have learned from studies of gaze allocation in natural vision. Clear principles of selection are found across many instances of natural vision and these are not the principles that might be expected from picture-viewing studies. We discuss the emerging theoretical framework for gaze allocation on the basis of reward maximization and uncertainty reduction.

Are fixations in static natural scenes a useful predictor of attention in the real world?

“…This is problematic for the many studies that wish to make conclusions about real world attention from static scene scanning, and it validates a central claim of Tatler et al, (2011) who criticised the picture-viewing paradigm. This key finding also differs from t 'Hart et al, (2009) who found that fixations in static frames (and particularly in continuous movies) could predict gaze in the real world better than spatial biases alone.…”

“…And one could do even -28-better than that by including where the eye is usually positioned in the head, and not being concerned at all about the content of any of the scenes. We began our study by discussing the work of Tatler et al (2011) who provided the sobering warning that looking behaviour in computer images may not be especially predictive of looking behaviour in real everyday life. The present study provides a test of that proposal and validates it.…”

“…Despite these advantages, several researchers have acknowledged the fact that most experimental data on this topic are derived from experiments with static images that are quite different from real scenes (Tatler et al, 2011;Henderson, 2007). This raises issues not dissimilar from those identified by Gibson (1979) and Neisser (1976) in their criticisms of artificial laboratory science.…”

“…However, Tatler et al (2011) argue that picture viewing is unlikely to tell us much about natural visual attention, and that a reliance on this paradigm has led to models of attention that over-emphasize pixel and feature-based guidance. Their critique points out that static scenes are typically smaller than our view of the real world, and that they do not replicate -5-dynamic cues.…”

Irregularity-based image regions saliency identification and evaluation

“…Area Under the Curve was used for this purpose in which the saliency map is converted to binary image and then the AUC is calculated and compared to the AUC extracted from the ground truth data (Lin, et al, 2013) (Gide & Karam, 2012), (Zhao & Koch, 2011), (Erdem & Erdem, 2013), and (Kim & Milanfar, 2013). Tatler supported the idea of empirical evaluation of the saliency since they were interested more in mimicking the natural behaviour of the humans' eyes (Tatler, et al, 2011). Precision and Recall, is another method was used by many publications such as (Fang, et al, 2012), (Stöttinger , et al, 2009), (Liu, et al, 2011), (Achanta, et al, 2009), and (Anon., 2013) to evaluate the saliency algorithms.…”