We first consider a noisy curve f which has jumps, corners and curved parts. In Figure 2, the results of applying our model on this noisy curve are listed by showing the function u and the residual f ° u at steady state. To make a comparison, we also list the results at the steady state given by the ROF model.
From the results in Figure 2, as expected, the oscillation noise is removed; the jumps, corners are all preserved. Besides these, in contrast to the ROF model, our model deletes the staircase phenomenon and also preserves image contrasts. More- over, the residual f ° u given by our model is more evenly distributed than the one by the ROF model, which suggests that our model is more efficient in noise removing.