Smith D K, Marks Ii R J
Appl Opt. 1981 Jul 15;20(14):2476-83. doi: 10.1364/AO.20.002476.
Digital implementations of a closed form 2-D bandlimited image extrapolation algorithm are presented for a number of elementary target images. The resulting figures of merit empirically verify the assumption that extrapolation is better near to where the image is known. Results of extrapolating a truncated image perturbed by white Gaussian noise suggest a required SNR of the order of 10(10).
针对一些基本目标图像，提出了一种封闭形式二维带限图像外推算法的数字实现方法。所得的品质因数通过实验验证了这样一个假设，即在外推靠近已知图像的区域时效果更好。对受白色高斯噪声干扰的截断图像进行外推的结果表明，所需的信噪比约为10(10)。
