Ant Colony Optimization Based Salient Object Detection for Weak Light Images
Nan Mu, Xin Xu, Xiaolong Zhang · 2018
Since substantial attention has been received over last decade, salient object detection becomes a fundamental research in computer vision community and has tremendous potential to solve real world problems. Although various works have achieved remarkable success in saliency detection tasks, there still remains a challenging issue on how to process weak light images, mainly due to low signal to noise ratios and limited effective features. In this paper, an ant colony optimization (ACO) based framework is proposed to detect the salient object presented in weak light backgrounds. The input image is first represented as an undirected graph, the nodes of which are generated by superpixel segmentation. Then, the optimal feature selection strategy is adopted to capture the useful information and the spatial contrast information is introduced to explore the global saliency cues. To acquire more accurate saliency estimation, the ACO measure is conducted to optimize the saliency map. Experiments on three public datasets and our weak light dataset demonstrate that the proposed model has promising performance compared to 11 state-of-the-art saliency approaches.