Learning from neighbours for saliency optimization

Saliency maps computed by each salient object detection algorithm show similar effect against ground truth. Therefore this paper proposes to optimize the saliency value of each pixel based on the saliency values of its neighbouring pixels. It proposes a machine learning-based method to learn the relationship between ground truth and saliency maps and among saliency values of neighbouring pixels. Then it applies the learned models to a new saliency map and then confines the fitted values to the legitimate range to obtain the optimized saliency map. The optimized saliency map is closer to the ground truth and better than the original saliency map. The experimental results show that the values of many saliency evaluation metrics improved significantly and some of them remained unchanged. They illustrate that the proposed algorithm is suitable for optimizing many salient object detection algorithms.