A supervised feature weighting method for salient object detection using particle swarm optimization

Salient object detection (SOD) is the task of localizing and segmenting the most conspicuous foreground object(s) from a scene. SOD has recently grabbed much attention in computer vision mainly because it helps find the objects that efficiently represent a scene and thus solve complex vision problems such as scene understanding. Since combining features is one of the most common ways to compute the final saliency map in SOD, considering the relative contribution of each feature is as important as feature extraction. In this paper, we develop a feature weighting method by utilizing Particle Swarm Optimization (PSO) to generate a suitable weight vector in order to combine features effectively. The performance of the new method is compared with six existing methods on three different data sets. The results suggest that by considering the importance of each feature in the combination process, the proposed method has achieved better performance than that of the competitive methods.