基于统计分析的图像去噪

Noise removal of image on the statistic analysis

噪声虽然降低了图像像素之间的相关程度,但是相关性仍然存在.本文以图像邻域像素块为出发点,运用图像邻域像素块之间的相关性分析图像的能量分布,进一步挖掘图像能量分布的特征向量,运用特征向量重构图像达到去除噪声的目的.本文挖掘的特征向量与邻域像素个数、像素块之间的相关性和特征向量的选取有关,从实验上讨论了这三个因素对去噪的影响,并与传统算法相比,本文算法对峰值信噪比(PSNR)较低的图像具有较强的噪声抑制能力,同时图像的边缘和纹理等细节保留较多.

The image noise reduces the degree of correlation between the image pixel and its neighbors, but the correlation still exists. As a staring point of neighbor pixel block, the paper uses the correlation between them to analyze the image energy distribution. In order to achieve the purpose of removing noise, the authors further mine the feature vectors of the energy distribution and use these feature vectors to reconstruction image. The feature vectors are relevant to the number of the neighbor pixels, the correlation between image blocks and the selection of feature vectors. The experiments analyze the impact of these factors on de-noising. Compared with traditional algorithms, the algorithm proposed in the paper has the strong capability to suppress noise and keeps more image details.