Salient Region Detection Using Local and Global Saliency

In this paper, we present a novel local-global salient region detection method. We first obtain the smoothed image via gradient minimization, resulting in more homogeneous background. Then, we partition the smoothed image into a set of regions and compute the region saliency by measuring the dissimilarity and spatial distance. Furthermore, we adopt the global color distribution, including the color coherence, to yield global saliency region. Finally, we combine the local-and-global salient regions and the composition information to obtain the overall salient regions. Experimental results show the efficacy of the proposed method in comparison with the existing methods.