DOI: 10.18178/wcse.2019.06.033
Super-Resolution for Mixed-quality Stereo Images based on Patch Matching
Abstract— We propose a novel method for mixed-quality stereo images super-resolution (SR) based on patch matching. Previous related methods always put effort on disparity estimation which cannot achieve the accuracy for SR. In this paper, we directly utilize the information from the high-resolution (HR) image to reconstruct the low-resolution (LR) image. More specifically, the vanishing point estimation algorithm is adopted to identify the planes correspondence in the pair of images. Then we search the best matching patch for LR from HR in the corresponding planar area. Furthermore, we define a curvature criterion that can keep the patches with high-frequency information during patch matching process. Compared with state-of-the-art methods, the proposed framework gains 1.39 PSNR improvement and 0.003 SSIM improvement.
Index Terms— Stereo images, super-resolution, patch matching.
Chengtao Cai, Bing Fan, Haiyang Meng
College of Automation, Harbin Engineering University, CHINA
Cite: Chengtao Cai, Bing Fan, Haiyang Meng, "Super-Resolution for Mixed-quality Stereo Images based on Patch Matching," Proceedings of 2019 the 9th International Workshop on Computer Science and Engineering, pp. 219-226, Hong Kong, 15-17 June, 2019.