Abstract
The more salient point of a scene, the more it attracts visual attention. A saliency-based visual attention model can simulate human eye movement. While saliency is computed from local information, ensemble is computed from spatially distributed objects. Human can also immediately get global information of the scene as ensemble information. However it remains unclear whether ensemble is used for visual search. The research questions are whether, when and how saliency and ensemble affect human search process. Our prior research suggests that eye movements depend primarily on saliency at first and ensemble gradually affects them after a few saccades. Gabor patches were placed as targets in the left and right side of the screen. The task was to count the point of Gabors distributed on pink-noised background and judge which side of the screen included just 20 points, during which their eye movements were recorded. In each trial, among white Gabors with 1 point, there was only one colored (the most salient) Gabor with 2 or more points, in the side with 20 points (congruent condition), or in the side with less than 20 points (incongruent condition) and the difference in points between the left and right sides were manipulated. Since the saliency of colored Gabor includes both physical and value components, the value of colored Gabor may be responsible for the higher saliency at the initial stage. To examine this possibility, the current experiment used all Gabors with equal point (1 point), and now there was no initial advantage of colored Gabor over ensemble information, in that the direction of the first saccade is not significantly different from chance level in the incongruent condition. It suggests that visual saliency and ensemble work simultaneously when the salient object has an equal value. This work was supported by JSPS KAKENHI grant #16H01506 and #16H01727.
Meeting abstract presented at VSS 2017