Edge Preserving Filter Selection for Noise Removal and Histogram Equalization

Enhancement of image contrast can be done efficiently and simply using the Histogram Equalization (HE) technique. Noise amplification, over-enhancement, inconsistency in brightness and other such drawbacks are faced during HE along with loss of structure information. A novel edge preserving filter-selection scheme is proposeding this paper to simultaneously overcome these drawbacks. Along with certain regularizations, the generic data-fitting components and variational energies are reduced using a filter-based approach. However, the total energy required by the system is maintained. Total variation, Mean Curvature, Bernstein and Gaussian Curvature filters are the fast discrete filters presented in this paper that ensures that energy variation regularizes. Filter selection is performed based on the output of the comparison algorithm. Noise amplification and over enhancement are avoided and both qualitative, as well as quantitative evaluation of the filters, is performed. The energy of the overall model can be reduced rapidly using these filters. The experimental results prove the efficiency of these filters in regularization models.