A fuzzy preservation-based total variation filter for removal of random-valued impulse noise

In this paper, a fuzzy preservation-based total variation filter (FPTV) is proposed for removal of random-valued impulse noise. In our decision-based denoising scheme, Adaptive Center Weighted Median Filter (ACWMF) is firstly ameliorated by us to employ the variable window size technique to improve its detection ability especially in highly corrupted images. The improved ACWMF not only judge a pixel noisy or not, but also render the confidence coefficient (CF) for each pixel to evaluate its potential to be an impulse. Then, a function is designed with noise level p and CF as its arguments to determine pixel-wise the tradeoff between the smoothness term and the data fidelity term in total variation energy functional. After minimizing the energy functional, we obtain the restored image. Extensive simulation results show that our proposed method outperforms some representative algorithms, both in vision and quantitative measurements like peak signal-to-noise ratio (PSNR) and mean absolute error (MAE).