Enhanced saliency prediction via free energy principle
- Publication Type:
- Conference Proceeding
- Citation:
- Communications in Computer and Information Science, 2019, 1009 pp. 31 - 44
- Issue Date:
- 2019-01-01
Closed Access
| Filename | Description | Size | |||
|---|---|---|---|---|---|
| Pages from 2019_Book_DigitalTVAndMultimediaCommunic.pdf | Published Version | 1.07 MB |
Copyright Clearance Process
- Recently Added
- In Progress
- Closed Access
This item is closed access and not available.
© Springer Nature Singapore Pte Ltd 2019. Saliency prediction can be treated as the activity of human brain. Most saliency prediction methods employ features to determine the contrast of an image area relative to its surroundings. However, only few studies have investigated how human brain activities affect saliency prediction. In this paper, we propose an enhanced saliency prediction model via free energy principle. A new AR-RTV model, which combines the relative total variation (RTV) structure extractor with autoregressive (AR) operator, is firstly utilized to decompose an original image into the predictable component and the surprise component. Then, we adopt the local entropy of ‘surprise’ map and the gradient magnitude (GM) map to estimate the component saliency maps-sub-saliency respectively. Finally, inspired by visual error sensitivity, a saliency augment operator is designed to enhance the final saliency combined two sub-saliency maps. Experimental results on two benchmark databases demonstrate the superior performance of the proposed method compared to eleven state-of-the-art algorithms.
Please use this identifier to cite or link to this item: