基于小波系数层间相关性的图像噪声方差估计

Noise variance estimation based on inter-scale correlation of image coefficient of wavelet transform

噪声的方差估计是含噪图像处理中的常见问题之一,其基本思想是通过某种方法寻找含噪图像中的"纯"噪声子图像来估计原噪声方差。传统方法是通过空域或频域采样,得到该子噪声图像,然后直接对其估计方差,它对图像信息的分布有要求。在传统频域采样方法的基础上,提出一种结合图像小波变换系数层间相关性的新方法。其过程是:对第一级有效小波分解的斜向子块进行分析,利用小波变换系数的层间相关性,去除其中的图像信息,得到更"纯"的子噪声块,再估计其方差。通过仿真实验和实际4 f系统输出图像实验证明,该方法比传统方法的估计结果更准确,更适合带宽较低的系统图像和图像本身高频信息较丰富的场合。

Variance estimation is a common problem in noised image processing. The basic idea is to get a sub-image that includes only ＂pure＂ noise to estimate the variance of original noise. The traditional way is to obtain the sub-image by sampling the noised image in spatial or frequency domain, then calculate its variance to replace that of the original noise. It has some requirement on the distribution of the image information. Based on the traditional frequency sampling method, a new method was proposed. Firstly, choose the first effective oblique high sub-image in wavelet transform domain as the sub-noise image. Then wipe off the useful image information in it by using the inter-scale correlation of wavelet coefficient to get more ＂pure＂ noise sub-image and estimate its variance. Experiments of simulation and real 4f system show that the new method is more accurate than the traditional one. It is more suitable for the images from low-pass bandwidth system or images full of details.