Automatic Detection of Salient Object Based on Multi-Features

Many existing detecting measures of salient object mostly depend on the information between the pixels and lack for the comprehension of salient object features. This paper presents an unsupervised algorithm for automatically extracting the interesting object. We firstly extract the junior feature as basic elements of saliency by a biologically plausible Gabor model, and then combine the understanding knowledge of the salient object to define the local feature, which is used to locate the salient regions further. Finally, a novel extracting scheme based on fuzzy threshold approach is proposed, which acts on the conspicuity map integrated by the junior and local features to locate the salient object. Apply the algorithm on various images for the simulation experiments, the satisfying results are obtained. That shows the algorithm is feasible.