应用正则化影响函数扩散模型的星图噪声滤波

由于恒星星图的噪声滤波对保持星点的边缘细节要求较高，本文以塔基（Tukey）扩散模型与改善的PM（Perona-Malik）模型为基础，提出了一种基于正则化影响函数扩散模型的星图噪声滤波方法。该方法通过导数算子提取边界点集，利用图像中原始像素和噪声像素的空间分布特性对图像进行噪声滤波处理，并通过给定边界条件恢复图像边缘。由于避免了方差稳定（VS）变换，该方法可以直接处理高斯噪声。对普通图像和添加高斯噪声星图进行了仿真测试，并与普通扩散函数算法进行了比较。实验结果表明：提出的算法表现出了较好的噪声滤波能力，同时有效地保持了特征图像的边缘。相对于普通扩散函数算法其平均绝对误差降低了13.6%，峰值信噪比平均提高了6.1%。得到的数据显示，本方法的滤波能力优于普通的扩散函数方法，特别适用于星图的噪声滤波处理。

Star image noise filtering based on regularization influence function diffusion model

As noise filtering of a star image has a high demand for reserving details of star edge, a new star map noise filtering method on a regularization influence function diffusion model was proposed based on Tukey diffusion model and modified Perona-Malik model. The boundary point set was extracted by a derivative operator and the map noise was processed by filtering with the space distribution characteristics of the original pixel and the noise pixel in the images. Moreover, the image edge was recovered by a given boundary condition. Due to avoiding Variance Stabilization(VS) transform,it could process the Gaussian noise directly. Simulation experiments on a common image and a star map with Gaussian noise show that this method has good capability of noise filtering and can effectively reserve the edges of feature images. Compared with common diffusion function algorithm, the average error is reduced by 13.6% and the Peak Signal to Noise Ratio(PSNR) is improved by 6.1%. Filtering performance of the proposed method is better than that of common diffusion function method, especially suitable for noise filtering of star maps.