In this paper, we propose a simple visual saliencydetection model based on spatial position of salient objects and background cues. At first, discrete wavelet frame transform (DWDT) are used to extract directionality characteristics for estimating the centoid of salient objects in the input image. Then, the colour contrast feature performed is to represent the physical characteristics of salient objects. Conversely, sparse dictionary learning is applied to obtain the background feature map. Finally, three typical cues of the directional feature, the colour contrast feature and the background feature are mixed to generate a credible saliency map. Experimental results verify that the designed method is useful and effective.
|Title of host publication
|Proceedings of the 2019 25th International Conference on Automation and Computing (ICAC)
|Institute of Electrical and Electronics Engineers Inc.
|Number of pages
|Published - 11 Nov 2019
|25th IEEE International Conference on Automation and Computing - Lancaster, United Kingdom
Duration: 5 Sept 2019 → 7 Sept 2019
|25th IEEE International Conference on Automation and Computing
|5/09/19 → 7/09/19
- discrete wavelet transform
- saliency detection
- background features
- position prior