# 超分辨率阅读笔记（一）

$z_{l}=SHy_{h}+v$ where $(v\backsim N(0,\sigma^{2}I)$

$SH\in R^{N_{l}\times N{h}}$，显然其行数小于列数，导致其逆不稳定，从而会产生无穷多解。现有的方法都是为了使这个逆稳定，例如Tikhonov regularization, robust statistics and Total-Variation, sparsity of transform coefficients, and all the way to example-based techniques that use training set of images as priors.