Image Adaptation and Dynamic Browsing Based on Two-Layer Saliency Combination

In recent years, image adaptation has attracted more and more attention during the evolution of the integration of broadcasting, Internet and telecommunications. The diversity of display devices requires images to be resized for optimal display on different terminals. In this paper, an image adaptation scheme and a dynamic browsing strategy are proposed, which are based on the visual attention model (VAM) with two-layer saliency optimization. First, a VAM is constructed by optimizing the image saliency according to the global and local layers by simulated annealing and the saliency of focus of attention (FOA) can be ranked by the obtained saliency map. Then an image adaptation scheme is designed based on the obtained saliency map. In the proposed adaptation scheme, each predominant FOA is modeled as a rectangular region, and the image is adjusted corresponding to the display terminal and the rank of FOA. Based on the principle of global features precedence, a dynamic browsing strategy is developed for browsing large images on small display devices. Experiments on saliency map show that the cross-layer saliency detection algorithm has advantages on detecting salient regions accurately and reflecting the shift of FOA implicitly, which are useful for image adaptation. In addition, subjective evaluation experiments demonstrate the proposed image adaptation scheme and dynamic browsing strategy are feasible and have promising applications in practice.