亚像元成像系统细胞神经网络插值方法研究

Research on interpolation for improving the subpixel spatial resolution based on cellular neural network

亚像元动态成像技术是实现卫星相机小型化及提高相机空间分辨率的有效方法。采用细胞神经网络(CNN)实现了对亚像元图像的实时插值,对插值的过程进行了详细的分析,并与基于亚像元图像的传统插值方法进行了对比。结果表明,亚像元成像系统细胞神经网络插值方法不仅可以节约系统资源,而且还可以提高系统的运行速度。通过仿真证明了该方法的正确性。

The subpixel dynamic imaging technique is one of the valid new methods to reduce the size of camera and improve its spatial resolution. The methods of three-point interpolation for the subpixel image are realized based on the CNN（cellular neural network）. The design of 3×3 dimension CNN input template, the methods of interpolation and boundary conditions are analyzed and it is compared with other kinds of methods. It is advantaged for saving the resource of system and improving the running rate of system. The validity of the conclusions is demonstrated by simulation.