Abstract: This paper presents an algorithm to estimate the Signal Noise Ratio(SNR) for a single image. It defines that the SNR as the ratio of the signal intensity of image and the standard deviation of noise. Based on decorrelation criterion, it can obtain the optimal Gaussian filter, which is used in the accurate estimation of the standard deviation of noise, and it can obtain the robust estimation of the signal intensity of image through the Gaussian filtered image with a large scale. Quantitative experiments demonstrate that the proposed algorithm has high accuracy and validity.

Key words: Signal Noise Ratio(SNR), Gaussian filtering, standard deviation, decorrelation criterion

摘要： 提出一种估计单幅图像信噪比的算法。定义单幅图像信噪比为图像信号强度与噪音标准方差的比值。基于去相关准则可得到最优的抑制图像结构的高斯滤波，从而准确地估计图像噪音的标准方差，利用大尺度高斯滤波下的图像，得到稳定的图像信号强度估计。量化实验表明，该算法估计单幅图像的信噪比准确有效。

关键词: 信噪比, 高斯滤波, 标准方差, 去相关准则

CLC Number: 

• TN911.73