在获取车辆图像过程中，有很多因素会影响车辆图像的质量，如摄像机与车辆之间的相对运动引起的运动模糊，会导致车牌识别系统识别出来的车牌是错误的甚至识别不出车牌。根据运动模糊图像的特点，分析了运动模糊图像的退化模型，提出了基于改进的点扩散函数估计和全变分( TV)正则化的综合图像复原算法。利用改进的Radon变换法进行粗略计算，得到运动方向的近似解；通过双谱模型进行粗略计算，得到运动长度的近似解。得到上述两项结果后，通过TV正则化方法的应用，复原出模糊的图像，并采用梯度追踪算法求解该问题。从全方位的角度出发，对通过不同的方法得到的复原图像的质量进行评估。实验结果表明，经过优化的综合算法，不但能够提供较为准确的运动模糊参数，而且通过其复原出的车辆图像效果也是比较令人满意的。%In the process of taking vehicle images,there are many factors which can influence the quality of the vehicle images,such as motion blur caused by relative motion between video cameras and vehicles. It leads to the results that the license plate to identify by rec-ognition system is wrong,and even that the system couldn’ t recognize the license plate. According to the characteristics of the motion blur image,the degradation model of motion blurred image is analyzed,and an image restoration algorithm is proposed based on point spread function estimation and TV regularization method. Improved Radon transform is used to estimate the motion blur direction,and bi-spectrum model is applied to evaluate the motion blur length. After estimating the point spread function,the TV regularization method is used to recover the blurred image,and the gradient tracking algorithm is utilized to solve the problem. It evaluates and analyzes the quality of restored images from the perspective of subjective and objective in this paper. The experiment shows that the improved comprehensive algorithm can estimate the blurred parameter of motion-blurred images accurately and obtain satisfied effect.
展开▼
