A Study Of Image Retargeting Based On Seam Carving

With the proliferation of display devices, adjusting an image to heterogeneous devices with different sizes and aspect ratios is becoming more and more attractive. The way or the process of changing the image size or scale is called image retargeting. Traditional image retargeting methods such as scale methods and crop methods are all have defects in various degrees, which can’t satisfy our needs of both image aesthetics and image integrality. Content aware image retargeting achieves a good result in experiment so it is one of the hottest topics recently.As one of the most popular content aware image retargeting method, seam carving is simple operation and high efficiency. This method calculates the image importance map firstly, and then employs a dynamic programming to find the optimal seams. A seam is an optimal 8-connected path of pixels on a single image. By repeatedly carving out of inserting seams in one direction we can change the aspect ratio of an image. Through in-depth research we found that this method has some limitations and will fail in some images. In this paper we present a cartoon-texture based image retargeting method to improve the existing method on salience detection with complex background and large scale structure protection. This method contains a new image importance map and a new seam criterion. We first decompose an image into a cartoon and a texture part. The higher order statistics on the cartoon part provide reliable salient edges. We construct a salient object window and a distance dependent weight to modify the higher order statistics. The weighted higher order statistics is regarded as the salience of every pixel. This image importance map protects salient object from distortion by seam carving. We also propose a new seam criterion based on the analysis of the disadvantage of the original seam criterion. The newly proposed criterion regards seams as an optimal path which will bring the least energy to the image which tends to spread seam uniformly in nonsallient regions and helps to preserve large scale geometric structures. We evaluate our method visually and objectively, and compare the results with related methods. Our method performs well in retargeting images with cluttered backgrounds and in preserving large scale structures.At last, based on the analysis of the seam carving method and the development weproposed a local based uniform scaling method. This algorithm firstly departs an image into salient region and background part. And then in order to retarget the image into the target size we employ different scale of uniform scaling in the important regions and the background. Experimental results show that, our method performs well in retargeting images and in preserving large scale structures.