Salient Region Detection via High-Dimensional Color Transform Jiwhan Kim, Dongyoon Han, Yu-Wing Tai, and Junmo Kim
In this paper, we introduce a novel technique to automatically detect salient regions of an image via high-dimensional color transform. Our main idea is to represent a saliency map of an image as a linear combination of high-dimensional color space where salient regions and backgrounds can be distinctively separated. This is based on an observation that salient regions often have distinctive colors compared to the background in human perception, but human perception is often complicated and highly nonlinear. By mapping a low dimensional RGB color to a feature vector in a high-dimensional color space, we show that we can linearly separate the salient regions from the background by finding an optimal linear combination of color coefficients in the high-dimensional color space. Our high dimensional color space incorporates multiple color representations including RGB, CIELab, HSV and with gamma corrections to enrich its representative power. Our experimental results on three benchmark datasets show that our technique is effective, and it is computationally efficient in comparison to previous state-of-the-art techniques.
[1] Jiwhan Kim, Dongyoon Han, Yu-Wing Tai, and Junmo Kim, "Salient Region Detection via High-Dimensional Color Transform and Local Spatial Support", IEEE Transactions on Image Processing, Vol. 25, No. 1, pp. 9-23, Jan. 2016. [2] Jiwhan Kim, Dongyoon Han, Yu-Wing Tai, and Junmo Kim, "Salient Region Detection via High-Dimensional Color Transform", CVPR, 2014. [PDF]
[Code] (4.0MB, zip file, Matlab version) (2015.11.11 updated) Note : Our code requires the VLfeat library, which can be downloaded at : http://www.vlfeat.org/.
1. Which dataset did you use for training?
2. The computational time is more longer than that is described in the paper.
3. From the Table 1, why global contrast term has 9 dimensions? according to the definition, it's just a scalar.
4. The dimension of the high-dimensional color is l = 4*11=44, but how is the 'Dim' in Table 2 related to the 11 different color channel representations?
5. According to the Table 2, it seems that the high-dimensional color space has 44 dimensions, but why the Figure 5 has only three?
6. What are the failure cases?
[2] C. Yang, L. Zhang, H. Lu, X. Ruan, and M.-H. Yang, "Saliency detection via graph-based manifold ranking", CVPR, 2013. [3] Q. Yan, L. Xu, J. Shi, and J. Jia, "Hierarchical saliency detection", CVPR, 2013. [4] F. Perazzi, P. Krahenbuhl, Y. Pritch, and A. Hornung, "Saliency filters: Contrast based filtering for salient object detection", CVPR, 2012. [5] X. Shen, and Y. Wu, "A unified approach to salient object detection via low rank matrix recovery", CVPR, 2012. [6] M. Cheng, G. Zhang, M. Mitra, X. Huang, and S. Hu, "Global contrast based salient region detection", CVPR, 2011. [7] Y. Zhai, and M. Shah, "Visual attention detection in video sequences using spatiotemporal cues", ACM Multimedia, 2006. |