Image Seam Carving Based on Content Importance and Depth Maps

In many occasions images are transferred from one device to another with a different display size. In such situations images need to be resized. Image seam carving algorithms attempt to change the size of images while preserving image salient contents with minimum distortion. For this purpose energy map production is an important factor that can help determining the important objects and reducing distortion. Existing methods have some limitations and disadvantage. Due to lack of proper energy map that take advantage of both saliency and depth these methods could not attain proper results. In this paper, we develop a seam carving method based on object importance map. By jointly considering gradient, saliency and depth maps, we develop an efficient energy map. Furthermore, we proposed an edge preservation algorithm to avoid passing of seams through salient pixels. By taking advantage of both object importance map and edge preservation, our method achieved better seam carving performance. Experimental results show that the proposed approach produces better images than previous depth-aware and content-aware seam carving methods. Moreover, our results show that we can maintain both geometrical structures and salient objects while reducing visual artifacts.