Research On Stereo Image Retargeting

The development of multimedia technology promotes the expansion and application of image processing and display.In order to meet different viewing needs,the size of image is adjusted for the best display in different devices.The simplest image resizing method is uniform scaling,but this method usually causes the deformation in important objects and affecting the viewing comfort.For optimal display,it is necessary to consider how to effectively avoid visual distortion while preserving important content and structure when changing the size of image,and this process is called image retargeting.Compared with traditional 2D images,stereo images can provide information of depth.However,maintaining the depth information in the image retargeting process introduces new challenges.During stereo image retargeting,it should minimize the distortion in visual as well as the distortion in disparity.Serious distortion will happen if the 2D image retargeting algorithm is directly used in stereo images as ignoring the disparity relationship between pixels in the image pair.Therefore,it is necessary to study the stereoscopic image retargeting method capable of holding the depth information in combination with the characteristics of the stereoscopic image.In this paper,a pixel fusion based stereo image retargeting method is proposed,which could adaptively retarget stereo images with flexible aspect ratios,simultaneously minimizing the deformation of important objects and preserving the depth.Firstly,the seams are selected based on dynamic programming,and the energy function simultaneously considers the seam selecting and seam matching.Then,according to the selected seams,a matching map which represents the matching relationship between pixels in left and right views is established.Finally,a region-based scaling factor allocation mechanism is proposed in consideration of the effect of different pixel variations on the global depth extent.Experimental results demonstrate that the proposed method achieves preferable qualities in both depth and shape preservation for stereo image retargeting.In this paper,the method of stereo image retargeting based on mesh warping is implemented.Based on the traditional grid deformation framework,the shape preserving constraint is established according to the saliency value of the grid to savethe important objects in the image.In addition,according to the disparity distortion function,the depth preservation constraint is established to control the disparity of the image.According to the optimization algorithm,the transformed mesh is solved to obtain the retargeted stereo image.Experimental results show that this method can well preserve the important objects and the depth of stereo image.