凝视成像降质模型的超分辨率重建

发布时间：2019-07-08 18:45

【摘要】：针对吉林一号视频03星凝视成像的特点,为提高其图像空间分辨率,解决光学系统分辨率不足的问题,提出一种视频卫星超分辨率重建的新算法——凸集中间映射。以主动观点分析视频卫星凝视成像特点,建立了基于凝视成像的图像降质模型。为求解该降质模型的逆过程,以凸集理论为基础,建立了基于中间降质过程的约束集和相应的点投影算子,通过点投影算子逐帧修正高分辨率图像灰度值,最终将重建的高分辨率图像约束于凸集的交集上。实验结果表明,该算法使图像分辨率提高近30%,克服了同类算法具有投影误差和重叠伪影的缺点,图像质量评价指标均优于所列其他算法,8帧重建得到收敛解,对不同清晰度的图像重建均具有可行性和稳健性。说明该算法适用于视频卫星图像超分辨率重建。
文内图片：

图片说明：图２主动观点降质模型Ｆｉｇ．２Ｄｅｇｒａｄｅｄｍｏｄｅｌｏｆａｃｔｉｖｅｖｉｅｗ
[Abstract]:In order to improve the spatial resolution of Jilin No. 1 video 03 star staring imaging and solve the problem of insufficient resolution of optical system, a new algorithm for super-resolution reconstruction of video satellite, convex set intermediate mapping, is proposed. The characteristics of video satellite staring imaging are analyzed from the active point of view, and the image quality degradation model based on staring imaging is established. In order to solve the inverse process of the degradation model, based on the convex set theory, the constraint set based on the intermediate reduction process and the corresponding point projection operator are established. The gray value of the high resolution image is modified frame by the point projection operator, and finally the reconstructed high resolution image is constrained to the intersection of the convex set. The experimental results show that the algorithm improves the image resolution by nearly 30%, and overcomes the shortcomings of the same algorithm, such as projection error and overlapping artifact. The image quality evaluation index is better than the other algorithms listed. The convergence solution of 8 frames reconstruction is obtained, which is feasible and robust for different definition image reconstruction. It is shown that the algorithm is suitable for super-resolution reconstruction of video satellite images.
【作者单位】：中国科学院长春光学精密机械与物理研究所;中国科学院大学;长光卫星技术有限公司;
【基金】：国家自然科学基金青年基金(61505203) 中科院青促会专项基于智能仪器理念的一体化光学微小卫星关键技术
【分类号】：TP751

【相似文献】