一种应用于车牌检测的改进暗原色去雾算法  

An Improved Dark Channel Prior Algorithm for License Plate Detection

作　　者：师翔[1] 李民政[1] 

机构地区：[1]桂林电子科技大学计算机科学与工程学院,桂林541004

出　　处：《现代计算机（中旬刊）》2016年第6期66-70,共5页Modern Computer

摘　　要：针对暗原色先验方法不适用于白色和高亮区域的问题,提出一种改进的暗原色去雾算法,使还原图像在车牌检测时的表现更好。首先从大气中雾构成的角度出发,变型化简大气散射模型,得到去雾物理模型。然后利用暗原色先验方法求得透射率,并对原方法中会失真的高亮区域进行透射率修正。最后利用去雾模型还原图像。实验证明,改进的算法去雾效果更好,而且能够改善原暗原色算法中的失真情况,提高车牌检测的准确性。Aiming at the problem existed in the original Dark Channel Prior algorithm in highlight area, proposes an improved Dark Channel Prior algorithm for better performance in detecting license plate. Firstly, from perspective of fog formation, simplify the atmospheric scattering model, and then the dehazing model can be got. Then, uses dark channel prior method to estimate the transmittance function, and modi-fies the function in the high light area of the picture. Finally, the foggy picture can be restored by using the dehazing model. Experiments show that this algorithm has better performance in dehazing, and overcomes the distortion situation, thus raises the accuracy in license plate detection.

关 键 词：暗原色 透射率 车牌检测 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]