Saliency Detection based on spectra Destiny Peaks Clustering
Ruochen Xia, Guangwei Wang, Xiang Cheng · IOP Conference Series Materials Science and Engineering · 2020
Abstract Saliency detection mapping the whole salient object by simulating the human visual system is one of the fundamental problems in computer vision. In this paper we proposed a novel method to map the spatial vector features of image points to spectra, and decompose an image into large scale perceptually homogeneous elements for efficient salient region detection, using spectra destiny peaks clustering. a multi-layer saliency mapping is built based the size of regions, The final saliency map is intergraded in a hierarchical model. This method can overcome the common problem in saliency detection that the detection accuracy could be adversely affected if salient foreground or background in an image containing small-scale high-contrast regions. The experimental results show that the proposed method outperforms are improved greatly.