Rεςεαρch

Automatic contrast enhancement by variational minimax optimization

Contrast enhancement increases visibility of dim object features and improves overall visual appearance of imagery. We propose a variational Minimax Optimization method (MiMO) that seeks to explicitly balance the image contrast enhancement and information distortion in the paradigm of histogram equalization. A generic energy functional is formulated that combines the cumulative distances between histogram bars and a convex regularization term. With easy parameter setup, the algorithm is computationally efficient. The MiMO method reveals hidden details by spreading out the central modes of the histogram more evenly over the dynamic range despite the presence of a strong bright or dark background.