为了有效利用合成孔径雷达(SAR)图像丰富的地物纹理信息以解决现有超分辨率算法对SAR图像进行超分辨时有限的性能,提出了一种基于纹理分解的联合优化回归器算法用于低分辨率SAR图像的超分辨任务。该算法通过对低分辨率SAR图像和重构的高分辨率图像在局部和全局视角下的联合优化并辅以图像纹理信息来有效生成高质量的高分辨SAR图像。基于多种评价方法,在模拟数据集和真实数据集上开展实验。实验结果表明:所提算法强于其他典型的超分辨率算法。

In order to effectively use the rich surface texture information of synthetic aperture radar (SAR) image for the purpose of solving the limited performance of existing SAR image-based super-resolution algorithms, a joint optimized regressor algorithm based on texture decomposition is proposed for the super-resolution task of low-resolution SAR image. The algorithm effectively generates high-quality high-resolution SAR images through joint optimization of low-resolution SAR images and reconstructed highresolution images from both local and global perspectives, supplemented by image texture information. Based on multiple evaluation methods, experiments on simulated data sets and real data sets are carried out. Experimental results show that the proposed algorithm is stronger than other typical super-resolution algorithms.

TN957.52

2020年江苏省高职院校教师专业带头人高端研修项目(团队访学)(2020TDFX003)