Balanced energy for content-aware image resizing

This paper presents an image resizing method called balanced seam carving (BSC). First, we propose a new energy map to take into account both the edge energy and the visual saliency during resizing. Second, based on analyzing the drawbacks of previous forward energy function, we propose a novel energy criterion to evaluate the diagonal artifacts in addition to the previous horizontal and vertical artifacts. We implement this new criterion using the dynamic programming and the graph cut methods. The experiments demonstrate that our BSC method can improve the visual quality of the retargeted image and the robustness of the seam carving method. Moreover, the improved method is simple to implement, and can be easily applied to many existing seam carving based image resizing methods.