摘要
针对含高斯白噪声图像的噪声估计问题,提出一种改进传统分块法的新型算法。该算法提出灰度级范围对部分噪声的抑制作用,并因此造成对偏亮或偏暗图像的噪声估计有严重的欠估计。所提算法从解决此问题着手,合理结合滤波法对噪声的粗略估计结果得出溢出灰度级的边界条件。改进后的分块法自适应地选取划分图像的窗口大小、筛选噪声未溢出的子块及求取标准差排序后的数学统计参数。该算法不仅适用于噪声估计中常用的经典图像,也适用于现实生活中常见的各种监控图像,且噪声估计的结果受图像细节影响很小,对具有不同尺寸、不同信噪比、亮度不均衡及含不同等级噪声等特征的图像均取得较优的估计结果。实验结果表明,该算法具有更普遍的适用性、更高的精度和更好的鲁棒性。
To estimate noise variance in a white Gaussian noise image, an improved block-based algorithm was proposed. The improved noise estimation approach put forward that the gray-level restrains some of the noise. When dealing with brighter or darker images, this phenomenon may cause serious underestimation. The proposed approach started with the key to underestimation, got the boundary condition of overflowing the gray-level by combining filter-based method. The improved block-based method selected window size for partition, sub-blocks without overflowing, mathematical proportion parameter self- adaptively. The approach both applied to the classical noise estimation images and surveillance images which were more common in daily life. The improved block-based method was hardly affected by image details, it performed well in images with different sizes, different Signal-to-Noise Ratio （SNR） or uneven brightness. The experimental result shows that the proposed algorithm possesses wider applicability, higher accuracy and better robustness.
出处
《计算机应用》
CSCD
北大核心
2014年第7期2014-2017,共4页
journal of Computer Applications
关键词
噪声估计
分块法
欠估计
灰度级溢出
自适应
noise estimation
block-based method
underestimation
gray-level overflow
self-adaptive
