零散点停车位智能管理系统中的车牌图像超分辨率重建技术研究

Research on the Super Resolution Reconstruction Technology of License Plate Image in Intelligent Management System of Scattered Point Parking Place

针对车位管理系统监控视频下车牌图像会出现低分辨率问题,为了提高车牌图像分辨率,改进一种超分辨率重建算法。首先利用图像序列之间的关联性,使用块匹配进行运动估计,筛选出合适的帧序列,利用MAP法将筛选出的图像序列重建出合适的中分辨率图像。最后将中分辨率图像结合经过MOD算法训练出的超完备字典,利用稀疏表示的方法,重建为HR图像。实验结果表明,该方法合理利用车牌图像间的关联性和先验信息,提高重建出的高分辨率车牌图像质量。

In order to improve the resolution of license plate image, proposes a super-resolution reconstruction algorithm. Firstly, uses the correlation between image sequences, uses the block matching to estimate the motion, and selects the appropriate frame sequences. Then the filtered image sequences are reconstructed into appropriate moderate resolution images by MAP method. Finally, the moderate resolution image is combined with the super-complete dictionary trained by MOD algorithm, reconstructs the HR image by sparse representation. Experimental results show that the proposed method makes full use of the correlation and prior information between license plate images and improves the quality of the reconstructed high-resolution license plate images.