中文说明： 基于实例的超分辨率深深连接传统 _分析_ 内插方法，因为他们分享同样的目标。传统的插值方法 (例如线性，双三次，和三次样条插值)，通过已知像素的平均值作为计算 HR 网格中缺失的像素，隐式地施加 _smoothness_ 事先。然而，自然图像经常呈现强烈的不连续性，如边缘和角，和因此产生振铃和模糊输出图像伪影的平滑度事先结果。因此 SR 的目标是要实现更好的结果，通过使用更复杂的统计先验。
English Description：
example-based super resolution is deeply connected with traditional _analytical_ interpolation, as they share the same goal. Traditional interpolation methods (e.g. linear, bicubic, and cubic spline interpolation ), by computing the missing pixels in the HR grid as averages of known pixels, implicitly impose a _smoothness_ prior. However, natural images often present strong discontinuities, such as edges and corners, and thus the smoothness prior results in producing ringing and blurring artifacts in the output image. The goal of SR is thus to achieve better results, by using more sophisticated statistical priors.