中文摘要：

在噪声图像去噪、分割等算法中,需要知道噪声的分布模型和统计参数,为此提出了一个新颖的噪声估计算法.首先计算输入噪声图像中每个图像块的方差和局部熵的综合值;然后将所有综合值按降序排列,按序取综合值对应的标准差值进行去噪;最后通过图像质量评价算法筛选出最终的噪声估计结果.该算法不需要对输入图像进行如滤波、小波变换等复杂的预处理过程,直接对输入图像进行一系列的数据处理就可以得到噪声方差值,方法简单易于实现,计算效率高并且具有良好的鲁棒性,同时也可以指导BM3D等去噪算法进行自适应去噪.

英文摘要：

In the de-noising and segmentation algorithm used to deal with the noise image, it is necessary to know the distribution model and the statistical parameters of noise. A novel noise estimation algorithm was thus proposed. First, the combined value of the input noise image variance and local entropy of each image block was calculated. Then all the comprehensive values were arranged in a descending order, and de-noising was calculated using the corresponding standards deviations in that order. Finally, final noise estimates were selected using the image quality evaluation algorithm. The proposed algorithm does not need pre-processing such as complex filtering, wavelet transform, etc. , and can obtain the variance of noise by directly processing a series of input image data. It is simple and easy to implement, has high computational efficiency, and enable BM3D and similar de-noising algorithm to denoise adaptively.