摘要： 利用小波系数的层内相关性原理，并结合广义高斯模型，提出一种自适应邻域的阈值去噪方法。该方法通过计算以待处理系数为中心的不同邻域内的相关度系数，选择相关程度最好的邻域。对该方法中选择的邻域尺寸进行统计，发现分解的层次越高，较大的邻域出现的概率越大，这有利于保护边缘信息。实验结果表明，该方法优于固定邻域及阈值改进的邻域阈值方法，是一种有效的去噪方法。

关键词: 图像去噪, 小波变换, 自适应滑窗, 贝叶斯阈值, 邻域系数, 相关系数

Abstract: Using intra-scale dependency of wavelet coefficients and generalized Gaussian model, this paper proposes an adaptive neighboring threshold value de-noising method. By calculating the relative coefficients in the different neighborhood, a well neighborhood of relative coefficient is choosen. Through gathering statistics of neighboring size choosen, decomposition level is higher, the probability is biger in a large neighborhood, thus makes for profecting information. Experimental result shows that the method is better than that of fixed neighborhood and improved threshold value, it is a valid de-noising method.

Key words: images de-noising, wavelet transform, adaptive slide windows, Bayes threshold value, neighboring coefficients, relative coefficients

中图分类号: 

• TP18