Scanpath estimation based on foveated image saliency
详细信息 查看全文
- 作者：Yixiu Wang ; Bin Wang ; Xiaofeng Wu ; Liming Zhang
- 关键词：Selective visual attention ; Saliency ; Gaze shift ; Eye movement ; Scanpath
- 刊名：Cognitive Processing
- 出版年：2017
- 出版时间：February 2017
- 年：2017
- 卷：18
- 期：1
- 页码：87-95
- 全文大小：
- 刊物主题：Neurosciences; Behavioral Sciences; Artificial Intelligence (incl. Robotics);
- 出版者：Springer Berlin Heidelberg
- ISSN：1612-4790
- 卷排序：18
文摘
The estimation of gaze shift has been an important research area in saliency modeling. Gaze movement is a dynamic progress, yet existing estimation methods are limited to estimating scanpaths within only one saliency map, providing results with unsatisfactory accuracy. A bio-inspired method for gaze shift prediction is thus proposed. We take the effect of foveation into account in the proposed model, which plays an important role in the search for dynamic salient regions. The saccadic bias of gaze shifts and the mechanism of inhibition of return in short-term memory are also considered. Based on the probability map derived from these factors, candidates for the next fixation can be randomly generated, and the final scanpath can be acquired point by point. By the evaluation of objective measures, experimental results show that this method possesses better performance in several datasets than many existing models do.