基于自适应耦合PDE模型的车牌图像去噪研究

Study on the License Plate Image De-noising Based on Adaptive Coupling PDE Model

摘要:

针对车牌识别预处理中的图像去噪问题,提出一种自适应耦合偏微分方程(PDE)去噪模型;该模型在各项异性扩散模型的基础上,构造一种新的去除椒盐噪声的扩散项,能够根据噪声图像特点自适应控制扩散速度,有效抑制椒盐噪声,并将新的扩散项与各向异性扩散模型进行耦合,并提出一种新的耦合系数计算方法,根据图像信息自适应计算耦合系数,使得新模型能够在新的扩散项和各项异性扩散模型间自适应转换,有效去除车牌图像中的混合噪声;为了抑制去噪引起的图像边缘模糊问题,引入振动滤波进行逆滤波,增强图像的边缘信息;实验结果表明,自适应耦合PDE模型能更有效去除车牌图像中的混合噪声,保护图像的边缘信息,提高图像的峰值信噪比(PSNR);去噪后的图像更有利于后续的字符分割与识别,有效提高车牌图像的识别准确率。

Abstract:

According to the image de-noising problem in the license plate recognition preprocessing, we propose an adaptive coupling PDE de-noising model. We construct a new diffusion term on the basis of the anisotropic diffusion model, which can control the diffusion velocity adaptively and inhibit the salt and pepper noise according to the features of the noisy image. We use the new diffusion term to couple with the anisotropic diffusion model. And we present a new method to calculate the coupling coefficient adaptively, so that the given model can transform between the new diffusion term and the anisotropic diffusion model adaptively and remove the mixed noise effectively. In order to inhibit the image edge blurring, we introduce in the shock filter to do reverse filtering and enhance the image edge. The simulation test results show that the new model can remove the mixed noise of license plate image effectively, preserve the image edge information better, and improve the PSNR. The de-noised image is beneficial to the subsequent character segmentation and recognition. The adaptive coupling PDE de-noising model can improve the correct rate of license plate recognition effectively.