随着双目成像技术的发展,双目图像超分辨在近年来得到了广泛关注。不同于单幅图像超分辨,双目图像超分辨可以利用左右视角图像间的互补信息进一步提升图像重建质量。对双目图像超分辨领域的最新进展进行了综述。首先,介绍了双目成像的基本理论。接着,对现有的双目图像超分辨算法和双目图像数据集进行归纳总结。随后,在基准数据集上对几种基于深度学习的主流算法的性能进行了评测,并探究了不同训练集对超分辨算法性能的影响。最后,总结了双目图像超分辨所面临的挑战,并展望了其未来的研究方向。

With the development of stereo imaging technology, stereo image super-resolution has attracted increasing attention in recent years. Different from single image super-resolution, stereo image super-resolution can further improve the quality of image reconstruction by using the complementary information between left and right view images. In this paper, we present a comprehensive survey on recent advances of stereo image super-resolution field. Firstly, we introduce the fundamental theory of stereo imaging. Then, we review the existing stereo image super-resolution algorithms and stereo image datasets. Afterwards, we comprehensively evaluate several mainstream deep learning-based stereo image super-resolution algorithms on benchmark datasets, and investigate the influences of different training sets on the performances of super-resolution algorithms. Finally, we summarize challenges and prospect future research directions in stereo image super-resolution.