Super-resolution reconstruction of low-resolution vehicle plates: A comparative study and a new algorithm

Traffic video surveillance cameras are widely used in China today. However, the resolution of the output video is very low and it is difficult tSuper-resolution image reconstruction techniqueso recognize the license number of the vehicles that broke traffic regulations. The Super-resolution image reconstruction techniques thus can be used to solve this problem. In this paper, we examine several important superresolution algorithms and developed a novel method. In the proposed algorithm, the dominant edges in the pre-reconstructed high-resolution image are first detected and back-projected onto the low-resolution image, edge-based segmentation is then applied to the observed low-resolution image. Each segmented region is extended and reconstructed separately. Finally, all the reconstructed regions are fused to get a single high-resolution image.