变分贝叶斯估计图像自适应滤波去燥算法

摘要：图像在拍摄获取、存储和传输等过程中，由于操作、光线等会造成各种噪音。变分贝叶斯估计图像自适应滤波去燥算法，通过引入新的隐变量，给出过程噪声后验分布；定义贝叶斯估计图像滤波模型，推导、计算真实分布和近似分布之间的Kullback-Leibler距离；以迭代递推方式估计目标状态，提高状态估计精度，实现图像自适应滤波去燥。实验表明该算法滤波去燥效果明显，能最大程度保护图像细节。

Abstract: In the process of image acquisition, storage and transmission, due to various noises caused by operation, light and so on.the adaptive filtering removing algorithm for image based on variational Bayesian estimation is proposed. By introducing new hidden variables, the posterior distribution of process noise is given. The image filtering model of Bayesian estimation is defined, and the Kullback-Leibler distance between the real distribution and the approximate distribution is deduced and calculated to iterate. Recursive method is used to estimate the target state, improve the accuracy of state estimation, and realize image adaptive filtering and removing. Experiments show that the algorithm has obvious filtering effect and can protect image details to the greatest extent. 

关键词：变分贝叶斯；估计；去燥算法