Share Email Print
cover

Proceedings Paper

Visual saliency detection based on modeling the spatial Gaussianity
Author(s): Hongbin Ju
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

In this paper, a novel salient object detection method based on modeling the spatial anomalies is presented. The proposed framework is inspired by the biological mechanism that human eyes are sensitive to the unusual and anomalous objects among complex background. It is supposed that a natural image can be seen as a combination of some similar or dissimilar basic patches, and there is a direct relationship between its saliency and anomaly. Some patches share high degree of similarity and have a vast number of quantity. They usually make up the background of an image. On the other hand, some patches present strong rarity and specificity. We name these patches “anomalies”. Generally, anomalous patch is a reflection of the edge or some special colors and textures in an image, and these pattern cannot be well “explained” by their surroundings. Human eyes show great interests in these anomalous patterns, and will automatically pick out the anomalous parts of an image as the salient regions. To better evaluate the anomaly degree of the basic patches and exploit their nonlinear statistical characteristics, a multivariate Gaussian distribution saliency evaluation model is proposed. In this way, objects with anomalous patterns usually appear as the outliers in the Gaussian distribution, and we identify these anomalous objects as salient ones. Experiments are conducted on the well-known MSRA saliency detection dataset. Compared with other recent developed visual saliency detection methods, our method suggests significant advantages.

Paper Details

Date Published: 13 April 2015
PDF: 6 pages
Proc. SPIE 9522, Selected Papers from Conferences of the Photoelectronic Technology Committee of the Chinese Society of Astronautics 2014, Part II, 95222N (13 April 2015); doi: 10.1117/12.2182055
Show Author Affiliations
Hongbin Ju, China State Shipbuilding Corp. (China)


Published in SPIE Proceedings Vol. 9522:
Selected Papers from Conferences of the Photoelectronic Technology Committee of the Chinese Society of Astronautics 2014, Part II
Xiangwan Du; Jennifer Liu; Dianyuan Fan; Jialing Le; Yueguang Lv; Jianquan Yao; Weimin Bao; Lijun Wang, Editor(s)

© SPIE. Terms of Use
Back to Top