Image Decomposition Model OSV with L 0 Sparse Regularization ?

Images decomposition attracts much attention in recent years. Many variational methods were proposed. The most famous variational decomposition model is OSV model because of its concise expression. However, the OSV model suffers from many drawbacks. One is that the discretization of the EularLagrange equation is hard because the 4th order term was introduced. The other drawback is that the effect of decomposition is weak. In this paper, we proposed a new decomposition model with L0 sparse regularization. The L0 sparse term is more sparse than traditional L1 norm. It is very useful in many image processing applications. The combination of OSV model and L0 regularization can improve the effect of traditional decomposition. To improve the efficiency of the minimization, the Split Bregman method for the improved OSV model is also designed. Experimental results validate the proposed model.