We propose a total variation-based variational model for nonblind binary image deblurring. The binary constraint is considered using the double-well function as the penalty term. We show the existence of a minimizer for the proposed model. By using o