A Generalized DAMRF Image Model for Super-Resolution of License Plates

In this paper, we propose a novel super-resolution image reconstruction algorithm to handle license plate texts in real traffic videos. A generalized discontinuity adaptive Markov random field (DAMRF) model is proposed based on the recently reported bilateral filtering, which is not only edge preservation but also robust to noise. Moreover, instead of looking for fixed value for the regularization parameter, a method for automatically estimating it is applied to the proposed model based on the input images. We use graduated non-convexity (GNC) optimization procedures to minimize the cost function. Results on synthetic and several real traffic sequences are presented, showing the effectiveness of the proposed method and demonstrating its superiority to the conventional DAMRF super-resolution method.