基于卡尔曼滤波的图像降噪方法研究

Image De-Noising based on Kalman Filter

为了改善噪声图像的质量,提出了一种基于KALMAN滤波的降噪方法,该算法采用递推性算法,因此,可以适用平稳与非平稳过程,这就解决了其他估计方法的限制性困难。该方法分析了噪声图像的特征,并且在一价高斯噪声的基础上改写了噪声图像的观测方程,同时,使用NSHP模型来构造图像的过程方程,大大的降低了卡尔曼滤波更新中的计算量。仿真结果表明,卡尔曼滤波方法可以明显的减弱了原始图像上噪声,并且有效的解决了图像滤波必然伴随的模糊细节问题,和其他传统噪声降噪方法比较,更好的保持了原图像中的一些线条,点和边缘的细节信息,体现了自己的自适应优点。

In order to improve the quality of noise image, a de-noising algorithm based on Kalman filtering is proposed. By adopting recursive algorithm, it could be applicable to the stationary process and non-sta- tionary process, thus solving the problem of limiting factors from other estimation methods. The character- istics of the noise image are analyzed, and then based on the first-order Gaussian color noise the equiva- lent observation of noise image is redefined. Meanwhile, NSHP （Non-Symmetric Half Plane） is applied to forming the process equation of image, thus considerably reducing the calculation complexity of updated Kalman filtering. Simulation results show that Kalman filtering could obviously reduce the noise of original image and effectively solve the fuzzy details resulted from image filtering. Compared with other traditional de-noising algorithms, the proposed method could better retain the image details, including lines, dots and margins and demonstrate its superiority of self-adaption.