Disparity-constrained resolution adaptation for stereoscopic images

Abstract

This paper addresses the problem of retargeting the stereoscopic images. We propose a disparity-constrained resolution adaptation method. In our method, we try to preserve the shapes of the prominent contents in the stereo images when they are displayed in the target resolution. Moreover, in order to improve the visual comfort, viewers can adjust the disparity of the stereo image pair linearly in our method. We first model the retargeting energy considering the images' significance map, content distortion and disparity constraints. Then we minimize the total energy and solve it as a nonlinear least square problem. The optimization is automatic without any post work, such as inpainting or occlusion handling. Finally, we demonstrate our method on the Middlebury stereo datasets and 3D animation screenshots. For stereo evaluation, the disparity map comparison results are provided. © Springer International Publishing Switzerland 2013.