基于隐含重起Arnoldi过程的参数估计

Implicitly restarted Arnoldi process based parameters estimation

提出在超分辨率复原中使用基于隐含重起Arnoldi过程来高效计算正则化参数的方法。通过隐含重起Arnoldi过程,可选择一个较好的初始向量。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。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 implicitly restarted Arnoldi process for the estimation of L-curve in super-resolution image restorations.The method can select the better initial vectors and generate orthogonal bases for the Krylov subspaces.The bases are small and condensed Hessenberg matrices which represent orthogonal projections of the large and sparse system matrix in super image restoration onto the Krylov subspaees.The method can reduce the computational complexity of the regular parameters.