1: \begin{abstract}

2: \noindent

3: %Intensity inhomogeneities in images constitute

4: %a considerable challenge in image segmentation.

5: Intensity inhomogeneities in images cause problems in gray-value based image segmentation since the varying intensity often dominates over gray-value differences of the image structures.

6: In this paper we propose a novel biconvex variational model

7: that includes the intensity inhomogeneities

8: to tackle this task.

9: We combine a total variation approach for 

10: multi class segmentation 

11: with a multiplicative model to handle the inhomogeneities.

12: In our model we assume that the image intensity 

13: is the product of a smoothly varying part and a component 

14: which resembles important image structures such as edges.

15: Therefore, we penalize in addition to the total variation of the label assignment matrix 

16: a quadratic difference term to cope with the smoothly varying factor.

17: A critical point of the resulting biconvex functional is computed by a modified

18: proximal alternating linearized minimization method

19: (PALM). We show that the assumptions for the convergence of the

20: algorithm are fulfilled.

21: Various numerical examples demonstrate the very good performance 

22: of our method. Particular attention is paid to the segmentation of 

23: 3D FIB tomographical images serving as a motivation for our work.

24: \end{abstract}

25: