标题:
基于Zernike 矩相似度联合的图像滤波Image Denoising by Zernike-Moment-Similarity Collaborative
Filtering
作者:
肖秀春, 赖剑煌
关键字:
图像去噪, 多边滤波, 非局域均值滤波, 相似性测度, Zernike 矩Image Denoising; Multilateral Filtering; Non-Local Means Filtering; Similarity Measure; Zernike Moments
期刊名称:
《Journal of Image and Signal Processing》, Vol.2 No.1, 2013-01-30
摘要:
非局域均值滤波(non-local means filtering, NLMF)采用图像块间灰度差测度像素间相似性，由于灰度差
易受噪声影响，这种相似性测度缺乏鲁棒性。图像块的Zernike 矩是块内像素灰度的统计量，且具有旋转无关特
性，能在抑制噪声的情况下较好地描述图像块特征。由图像块的各阶Zernike 矩差代替灰度差可定义Zernike 矩
相似度；联合各阶Zernike 矩相似度经加权平均可估计出所处理像素的灰度。仿真实验及分析表明文中算法相比
直接采用灰度差定义相似度的算法，能更好地去除噪声，获得更高的峰值信噪比(PSNR)。
Because similarity function defined in non-local means filter is subject to image noise, it cannot robustly
represent the real similarity between pixels. Zernike moments are good statistics of the pixels in image patch, and have
rotation-invariant feature, so they can be utilized to describe image feature while resistance to noise. In this paper,
Zernike-moment-similarity is defined according to the difference of Zernike moments instead of pixel intensity, and
then the intensity of the processed pixel is estimated by weighting the intensities of the local window according to the
collaborative Zernike-moment-similarity. Simulation experiment results and analysis demonstrate that the presented
algorithm can achieve better performance and higher PSNR than the current algorithms which directly adopting intensity
difference as its similarity function.