Salient region detection combining spatial distribution and global contrast

Abstract

We present a novel salient region detection scheme combining spatial distribution and global contrast. Our scheme considers not only global contrast between different colors/regions, but also spatial relationships within the same color/region. The variance of the spatial position within the same region/color is used to measure spatial saliency. And we incorporate spatial saliency and global contrast saliency into the final saliency map through linear combinations. Experiments show that the proposed scheme not only performs better than five state-of-the-art methods on the publicly available data set, but it also is simple, easy to implement and efficient.