用优化方法提升模糊图像分辨率，尤其适合有运动的场景。
Super Resolution image reconstructs via total variation-based image deconvolution: a majorization-minimization approach
- 基于变分正则化与极大极小算法，提升超分辨重建质量。
- 在模拟数据中成功恢复出清晰图像，有效处理运动模糊。
- 适合科研人员研究图像重建，不适用于实时应用。
本文针对具有平面投影运动的图像序列，研究基于总变差正则化的超分辨率重建方法。首先建立超分辨率成像观测模型，采用插值与融合估计及凸集投影技术；通过Horn-Schunck算法计算图像序列的光流以估计运动。随后提出一种基于极大极小策略的总变差正则化方法，实现更优的重建结果。文中还探讨了从运动测量中进行超分辨率复原的问题。数值实验表明该方法能有效恢复清晰图像，尽管无法实现实时处理，但为后续研究奠定了基础。
原文摘要 · Abstract (English)
This work aims to reconstruct image sequences with Total Variation regularity in super-resolution. We consider, in particular, images of scenes for which the point-to-point image transformation is a plane projective transformation. We first describe the super-resolution image's imaging observation model, an interpolation and Fusion estimator, and Projection on Convex Sets. We explain motion and compute the optical flow of a sequence of images using the Horn-Shunck algorithm to estimate motion. We then propose a Total Variation regulazer via a Majorization-Minimization approach to obtain a suitable result. Super Resolution restoration from motion measurements is also discussed. Finally, the simulation's part demonstrates the power of the proposed methodology. As expected, this model does not give real-time results, as seen in the numerical experiments section, but it is the cornerstone for future approaches. Finally, the simulation's part demonstrates the power of the proposed methodology. As expected, this model does not give real-time results, as seen in the numerical experiments section, but it is the cornerstone for future approaches.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。