Zhang Liangpei, Yuan Qiangqiang, Shen Huanfeng, Li Pingxiang
The State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, China.
J Opt Soc Am A Opt Image Sci Vis. 2011 Mar 1;28(3):381-90. doi: 10.1364/JOSAA.28.000381.
Super-resolution image reconstruction, which has been a hot research topic in recent years, is a process to reconstruct high-resolution images from shifted, low-resolution, degraded observations. Among the available reconstruction frameworks, the maximum a posteriori (MAP) model is widely used. However, existing methods usually employ a fixed prior item and regularization parameter for the entire HR image, ignoring local spatially adaptive properties, and the large computation load caused by the solution of the large-scale ill-posed problem is another issue to be noted. In this paper, a block-based local spatially adaptive reconstruction algorithm is proposed. To reduce the large computation load and realize the local spatially adaptive process of the prior model and regularization parameter, first the target image is divided into several same-sized blocks and the structure tensor is used to analyze the local spatial properties of each block. Different property prior items and regularization parameters are then applied adaptively to different properties' blocks. Experimental results show that the proposed method achieves better performance than methods with a fixed prior item and regularization parameter.
超分辨率图像重建是近年来的一个热门研究课题，它是一个从移位的、低分辨率的、退化的观测数据中重建高分辨率图像的过程。在现有的重建框架中，最大后验（MAP）模型被广泛使用。然而，现有方法通常对整个高分辨率图像采用固定的先验项和正则化参数，忽略了局部空间自适应特性，并且求解大规模不适定问题所带来的巨大计算负荷是另一个需要注意的问题。本文提出了一种基于块的局部空间自适应重建算法。为了减少巨大的计算负荷并实现先验模型和正则化参数的局部空间自适应过程，首先将目标图像划分为几个大小相同的块，并使用结构张量分析每个块的局部空间特性。然后，针对不同特性的块自适应地应用不同的特性先验项和正则化参数。实验结果表明，所提出的方法比具有固定先验项和正则化参数的方法具有更好的性能。
