新颖的频域图像超分辨重建方法

New super-resolution reconstructing algorithm in image frequency domain

图像频域解混叠超分辨算法要求输入低分辨率图像帧数及其帧间亚像元位移满足一定的条件,因此限制了其应用范围。利用单帧超分辨技术和再采样函数从每帧输入低分辨率图像产生16帧相同分辨率的图像,再从中挑选满足限制条件的图像帧,从而消除了原算法要求的两个主要限制条件,建立了改进的二至多帧输入图像频域解混叠超分辨算法。三组仿真实验结果表明,该算法消除了输入帧数和帧间亚像元位移的限制条件,并且可以使峰值信噪比提高5dB左右。

The de-aliasing Super-Resolution （SR） algorithm in image frequency domain demands some limits on frame numbers and sub-pixel shifts between frames of input Lower Resolution （LR） images, which limits the application range of this algorithm. Using single frame super-resolution method and resample function, the authors produced 16 frames images which have the same resolution as input images from each input LR image and then selected images that meet the requirements from these produced images. Therefore, a novel de-aliasing SR algorithm in image frequency domain from two to many frames of input LR images was proposed. Three simulation experimental results indicate that the proposed algorithm removes the limits on frame numbers and sub-pixel shifts between frames and makes the PSNR of SR images to be increased by about 5 dB.