Content-aware Image/Video Retargeting In Arbitrary Resolution

For the content-aware media retargeting system, the most challenging problem ishow to balance the content loss and the deformation of the media. A recent study sug-gests that in most cases, viewers prefer sacrifcing content over inserting deformationto the media.Basedontherecentlyprioritizedobjectivesandthestatisticalfndings,weproposea novel media retargeting approach which considers content loss and deformation in avisual friendly manner.To implement the proposed image retargeting scheme, frst, we divide an inputimage into several strips according to the average strip energy density and the linestructure information. Then, we fnd the maximum resizing lengths which makes thedeformation acceptable and choose the proper retargeting operator for each strip ac-cording to the energy information. The process above will be repeated iteratively untilacroppingwindowcanbeselectedwhichmakesthecontentlossacceptable. Afterthat,we extend the proposed method to video retargeting which can protect spatial saliencyand temporal coherence via some dynamic programming methods.The experimental results show that our proposed method can protect more im-portant content information than traditional cropping and make less deformation thanprevious content-aware media retargeting method.