An improved model of double-entry adaptive TV

This paper describes the classic total variation de-noising model and its improved models, and gives a simple comparison of the characteristics of each model. An adaptive spatial fidelity term is intended to ease the smoothing effect by the second-order nonlinear filtering over the details. An adaptive regularization term is used to reduce the “staircase effect”, and to achieve a more stable and converged value. We proposed two improvements, one is a new adaptive regularization term based on local information of the pre-image; the other is a double-entry adaptive model. We use SNR to quantify the effect of de-noising, and residual images to evaluate the loss of details and texture of the images. The results show that the improved method can achieve good results in the case of high noise as well. Compared to the original method, the proposed has better noise robustness.