车牌图像色彩迁移与正则化约束去雾算法

针对雾天车牌图像模糊、车牌识别率低的问题，给出了车牌图像色彩迁移与正则化约束去雾算法。算法主要包含色彩迁移去雾和文本修复两个模块。采用MKL（Monge-Kantorovitch Linear Colour Mapping）色彩迁移算法，恢复雾天车牌颜色信息实现去雾；利用车牌的文本像素的强度和梯度特征对车牌图像进行正则化约束，实现车牌中文本的修复。实验结果表明，无论针对合成车牌雾图还是自然车牌雾图，去雾效果良好，且在薄雾、中等雾及浓雾三种不同雾度环境下都能够有效提高车牌识别率。

Defogging Algorithm Based on Color Transfer and Regularization Constraints of License Plate Images

Aiming at the problems of foggy license plate images and low license plate recognition rate in foggy weather, a defogging algorithm based on color transfer and regularization constraints of license plate images is proposed. The algorithm mainly includes two modules for defogging by color transfer and text repair. Firstly, the MKL color transfer algorithm is used to restore the color information of foggy license plate. Then, the intensity and gradient characteristics of the license plate’s text pixels are applied to regularize the license plate image, thereby repair text. Experimental results show that, regardless of the synthetic foggy license plate map or natural foggy license plate  map, the algorithm has a good defogging effect, and can effectively improve the license plate recognition rate in the three foggy environments of mist, medium fog and dense fog.