基于奇异值分解的Speckle噪声滤波算法研究

Study on Filter Algorithm for Speckle Noise Based on Singular Value Decomposition

Speckle噪声的复杂性和真值图像的不可知性增加了消噪算法设计的难度,传统算法在设计时借助污染像素估计灰度真值,同时忽略了位置信息对灰度估计的影响。借助奇异值分解和曲面拟合进行了消噪算法的研究与设计,首先在局部窗口进行奇异值分解,并对酉矩阵采用中值滤波进行消噪,其次引入参数修正奇异值以替代传统的阈值处理方法,最后在局部区域利用灰度曲面拟合技术重构消噪图像。实验表明,与传统的均值滤波、中值滤波相比,该算法能获得较为满意的消噪结果,具有较高的峰值信噪比和边缘保留指数。

The complexity of Speckle noise and the fact that the original image is unknowable brings filtering algorithms much difficulty. Traditional algorithms use the contaminated pixels to estimate the original ones and omit the inffluence from the location information. Based on singular value decomposition（SVD） and surface fitting, an denosing algorithm is designed in this paper. Firstly, SVD is operated on local area in the noise image, and median-filter method on unitary matrix. Secondly ,a parameter is introduced to revise the singular values to replace of traditional threshold method. Finally ,gTey surface fitting is used to restrict the original image. Experiments show that, comparing to the traditional mean - filter and median - filter, the proposed ＂algorithm can get a more satisfactory result and a higher Peak Signal to Noise Ratio（PSNR） and Edge Preserve Index（EPI）.