1: \begin{abstract}

2: We study inverse problems $F\left(\sol\right) =\data$ 

3: with perturbed right hand side $\gobs$ corrupted by so-called impulsive 

4: noise, i.e.\ noise which is concentrated on a small subset of the domain of 

5: definition of $\data$. It is well known that Tikhonov-type 

6: regularization with an $L^1$ data fidelity term yields significantly more 

7: accurate results than Tikhonov regularization with classical $L^2$ data 

8: fidelity terms for this type of noise.  

9: The purpose of this paper is to provide a convergence analysis explaining this 

10: remarkable difference in accuracy. Our error estimates significantly 

11: improve previous error estimates for Tikhonov regularization with 

12: $L^1$-fidelity term in the case of impulsive noise. 

13: We present numerical results which are in good agreement with the predictions 

14: of our analysis. 

15: \end{abstract}