Location and Super-Resolution Enhancement of License Plates Based on Video Sequences

Aiming at improving readability of the license plates in some real traffic surveillance video, a system of location and super-resolution of license plates is presented. After modeling the background by Gaussian mixture model and getting the vehicle region from the video, the license plates are located by connected component analysis and tracked by template matching. Then, the super-resolution reconstruction of the image sequences can be realized with projection onto convex sets (POCS) algorithm. The experiment suggests that it can effectively locate the license plates in the video, and an enhancement of the license plate image is obtained after super-resolution reconstruction.