Projection method for fractional Lavrentiev regularisation method in Hilbert scales (Q6040588)

This paper studies an ill-posed operator equation characterized by a linear bounded operator operating between two real Hilbert spaces. The source term of this equation is subjected to contamination by noise. Adopting a common approach, the investigation of the operator equation unfolds within an optimization framework. The impetus for this study originates from the smoothing properties of Tikhonov regularization. In this work, the authors study the fractional Lavrentiev regularization method, renowned for mitigating the excessive smoothing effects associated with Tikhonov regularization. The study is conducted within the framework of Hilbert scales. The authors furnish a comprehensive convergence analysis for the proposed method and derive an optimal order error estimate contingent upon a Holder-type source condition and a strategic parameter selection. Furthermore, a comparative analysis between the standard Lavrentiev method and the fractional Lavrentiev regularization method is presented. Numerical demonstrations in the realms of ill-posed integral equations, signal processing, and image restoration illustrate the effectiveness and viability of the developed framework.