Image retargeting effectively resizes images by preserving therecognizability of important image regions. Most of retargeting methods rely ongood importance maps as a cue to retain or remove certain regions in the inputimage. In addition, the traditional evaluation exhaustively depends on userratings. There is a legitimate need for a methodological approach forevaluating retargeted results. Therefore, in this paper, we conduct a study andanalysis on the prominent method in image retargeting, Seam Carving. First, weintroduce two novel evaluation metrics which can be considered as the proxy ofuser ratings. Second, we exploit salient object dataset as a benchmark for thistask. We then investigate different types of importance maps for thisparticular problem. The experiments show that humans in general agree with theevaluation metrics on the retargeted results and some importance map methodsare consistently more favorable than others.
展开▼
