Abstract

In this paper, we introduce a novel technique to au-tomatically detect salient regions of an image via high-dimensional color transform. Our main idea is to repre-sent a saliency map of an image as a linear combinationof high-dimensional color space where salient regions andbackgrounds can be distinctively separated. This is basedon an observation that salient regions often have distinctivecolors compared to the background in human perception,but human perception is often complicated and highly non-linear. By mapping a low dimensional RGB color to a fea-ture vector in a high-dimensional color space, we show thatwe 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 highdimensional 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.