摘要
在低分辨率视频序列的车牌识别中,针对序列中车牌图像分辨率低、噪声污染严重的问题,提出了一种基于超分辨率重建技术增强车牌图像的方法。在图像实现亚像素级配准的基础上,根据局部图像明显的结构信息,构建归一化卷积的局部结构自适应高斯核函数,并将不同序列中包含的不同车牌信息的图像融合成一幅高分辨率图像。实验结果表明,该算法与传统方法相比,重构出的高分辨率图像具有更高的图像信噪比,且边缘保持性更好,能够有效地重构出高分辨率的车牌图像,提高车牌识别的准确率。
In the license plate recognition of low-resolution video sequences,aiming at the problem of low license plate image resolution and serious noise pollution in the sequence,a method of enhancing license plate image based on super-resolution reconstruction is proposed. Based on the sub-pixel collocation of images, a local structure-adaptive Gaussian kernel function with normalized convolution is constructed according to the obvious structure information of the local images,and the images of different license plate information contained in different sequences are merged into one high resolution image. The experimental results show that compared with the traditional method, the reconstructed high-resolution image has higher image signal-noise ratio and better edge retention, and the high-resolution license plate image effectively can be reconstructed. The accuracy of license plate is enhanced.
作者
山显响
刘云清
SHAN Xianxiang;LIU Yunqing(School of Electronics and Information Engineering,Changchun University of Science and Technology,Changchun 130022)
出处
《长春理工大学学报（自然科学版）》
2018年第3期106-110,共5页
Journal of Changchun University of Science and Technology(Natural Science Edition)
关键词
车辆牌照
亚像素级配准
归一化卷积
超分辨率
license plate number；sub-pixel-level registration；normalized convolution；super-resolution
