基于ARNOLDI过程的参数估计

Arnoldi Process-based Parameter Estimation

本文提出在超分辨率复原中使用基于Arnoldi过程来高效计算正则化参数的方法。通过Arnoldi过程分解,该方法将大型稀疏系统矩阵投影到Krylov子空间上并表达成一个小型稠密的Hessenberg矩阵。给出了利用Hessenberg矩阵简化超分辨率复原中解计算的公式。推导了快速计算L曲线的定理。该方法可减少正则化参数的计算代价。

The paper proposes an efficient method based on the Arnoldi process for the estimation of L-curve in superresolution image restorations. Through the Arnoldi process the method can generate orthogonal bases for the Krylov suhspaces and small and condensed Hessenherg matrices which are matrix representations of the orthogonal projections of the large and sparse system matrix in super image restoration onto the Krylov subspaces. The paper presents the simple solution in super image restoration by the Hessenberg matrix and formulates the the theorem for quickly computing L-curve. The method can reduce the computational complexity of the regular parameters.