Generalised Gradient Vector Flow for Content-Aware Image Resizing
Image retargeting is devoted to preserve the visual content of images with a proper resizing, removing vertical and/or horizontal paths of pixels which contain low semantic information. In this paper, a method based on the Generalised Gradient Vector Flow (GGVF) is presented. The GGVF formulation allows the balancing of the smoothing term and data term of the flow by proper parameter tuning. The proposed approach has been tested by considering a data set of 1000 images and varying the percentage of resizing from 10% to 50% and for different values of the aim involved parameter K. Results show that our algorithm better preserves the important information compared to GVF and Seam Carving approaches. Preliminary results show an underlying relation between parameter K and the percentage of resizing has been also exploited.
KeywordsImage resizing Image retargeting Seam carving GGVF
- 3.Cho, D., Park, J., Oh, T.H., Tai, Y.W., Kweon, I.: Weakly- and self-supervised learning for content-aware deep image retargeting. In: ICCV, pp. 4568–4577, October 2017. https://doi.org/10.1109/ICCV.2017.488