基于非局部相似性的立体图像超分辨率技术

Super Resolution for Stereo Image Based on Nonlocal Similarity

针对透视几何,提出了一种新的基于非局部相似性的立体图像超分辨率算法。算法将混合分辨率技术中视点的下采样过程建模成退化模型,并把相邻的高分辨率视点图作为参考图。利用参考图及深度信息,合成一幅目标视点的高分辨率投影图作为初始估计。将非局部相似性作为正则项添加到退化模型中。在寻找相似子块时,对于公共视野区中的块,从参考图中寻找;而对于遮挡区的块,则从投影图中寻找。最后,用梯度下降法求最优解。仿真实验分别对合成图和真实图进行了测试,结果表明,该算法较现有的算法具有更好的视觉效果和峰值信噪比。

Combined with the stereo geometry,a new super resolution algorithm for stereo image is proposed based on nonlocal（ NL） similarity. The downsampling processing of low-resolution view in the mixed-resolution coding was modeled as the degraded imaging model,and the adjacent views were considered as the reference images. A high resolution projected view was synthesized as initial estimation using the reference images and depth information. Then we introduce the NL similarity regularization term into the degraded model. When searching for the similar patches,we search patches in visible areas from the reference images,while search patches in occluded areas from the projected image. At last we used the gradient descent algorithm to find the optimal solution. Tests for synthetic and real images are presented,and results show that the proposed algorithm has better visual appearance and higher peak signal to noise ratio（ PSNR） than the state-of-the-art algorithms.