Directional Total Generalized Variation Regularization
For image restoration tasks with directional textures, this method offers improved regularization, but it is an incremental extension of existing TGV.
This paper proposes a directional total generalized variation (DTGV) regularization for image restoration, particularly for images with textures following one main direction. Experiments show significant improvement in texture preservation and noise removal compared to total variation and TGV.
In inverse problems, prior information and a priori-based regularization techniques play important roles. In this paper, we focus on image restoration problems, especially on restoring images whose texture mainly follow one direction. In order to incorporate the directional information, we propose a new directional total generalized variation (DTGV) functional, which is based on total generalized variation (TGV) by Bredies \textit{et al}. [SIAM J. Imaging Sci., 3 (2010)]. After studying the mathematical properties of DTGV, we utilize it as regularizer and propose the L$^2$-DTGV variational model for solving image restoration problems. Due to the requirement of the directional information in DTGV, we give a direction estimation algorithm, and then apply a primal-dual algorithm to solve the minimization problem. Experimental results show the effectiveness of the proposed method for restoring the directional images. In comparison with isotropic regularizers like total variation and TGV, the improvement of texture preservation and noise removal is significant.