Analysis of the Robustness of Iterative Restoration Methods with Respect to Variations of the Point Spread Function

This paper presents an empirical comparison of the Landweber iterative image restoration method with the direct CLS method, with the goal of determining which is more robust against errors in the estimated point spread function. 1. Image Formation Process 2. Restoration Methods q Constrained Least Squares (direct) ( ) T T T cls H C C H H h 1 − + = a q Landweber (iterative) ) ˆ ( ˆ ˆ * 1 k k k f H g H f f − + = + b q Landweber with Regularization ( ) ( ) k k f C C H H I g H f ˆ ˆ * * * 1 a b b + − + = + q Landweber with Regularization and Deterministic Constraints ( ) ( ) { } k k f C C H H I g H P f ˆ ˆ * * * 1 a b b + − + = + where α is the regularization parameter, β is a convergence parameter, and C is a Laplacian operator. 3. Quality Measure [ ] [ ]      