Multiple-cue saliency measurement and optimized image composition for image retargeting

摘要

We present a novel image retargeting approach for high-definition displaying based on multiple-cue salient region extraction and optimized image composition, which allows resizing an image into user-specified aspect ratios while preserving prominent visual salient image information. Firstly, we sum up six properties for measuring the salient objects and apply them to region extraction. Then, according to our optimized rules of image composition, we evaluate the aesthetic scores with different extracted sub-images. Finally, we sort out the target image with the highest aesthetic score and retarget it to fit the final scale of displaying. The experimental results show the performance and effectiveness of our approach.