改进SA方法的卫星视频图像超分辨率重建

发布时间：2018-05-13 10:42

  本文选题：SA方法 + 卫星视频　； 参考：《测绘科学技术学报》2017年01期

【摘要】：为了提高卫星视频图像的质量,针对SA方法中无法实现小数倍超分重建和"黑色格网"问题,提出了改进的SA超分辨率重建方法。对于小数倍超分问题,首先建立参考帧低分辨率图像坐标系,以超分倍数构建参考帧高分格网,按表达的物方元素一致的原则,把所有低分辨率图像和高分辨率图像的坐标系统一在一个参考坐标系下;对于"黑色格网"改善问题,在使用运动估计参数时,为了减少由于运动补偿造成的误差,取消传统SA方法中四舍五入的运动参数补偿方法,改为以待求点为中心确定容许误差,通过容许误差选择参估点,用参估点计算高分辨率网格像素,以减少"黑色格网"现象。利用仿真和真实的卫星视频数据实验验证了该方法的有效性。
[Abstract]:In order to improve the quality of satellite video images, an improved SA super-resolution reconstruction method is proposed to solve the problem that the fractional overdivision reconstruction and "black grid" can not be realized in SA method. For the decimals overdivision problem, the reference frame low-resolution image coordinate system is first established, and the reference frame high-division grid is constructed by the super-split-multiple, according to the principle of consistency of the expressed object elements. Coordinate systems for all low-resolution and high-resolution images are unified in one reference coordinate system. For the improvement of "black grid", in order to reduce the error caused by motion compensation, when using motion estimation parameters, Instead of the rounding motion parameter compensation method in traditional SA method, the allowable error is determined by taking the point to be sought as the center, and the parameter estimation point is used to calculate the high-resolution grid pixels through the allowable error, so as to reduce the "black grid" phenomenon. The effectiveness of the proposed method is verified by simulation and real satellite video data experiments.
【作者单位】： 辽宁工程技术大学测绘与地理科学学院;
【基金】：国家自然科学基金项目(41501504) 2016年辽宁省教育厅一般项目(LJYL011)
【分类号】：TP751
，

本文编号：1882817