摘要
基于直方图的模糊滤波方法对图像的拖尾噪声去噪会导致图像模糊、残留的噪声较多等问题,本文提出一种新的基于广义高斯混合模型的图像去噪方法.首先,建立图像的广义高斯分布及其有限混合模型;其次,通过像素周围点特征值的变化范围确定噪声数据;最后,利用广义高斯函数构建一个加权平均滤波器进行图像去噪.对基于直方图的滤波方法、经典的偏微分方程和本文方法进行比较实验,结果表明本文方法具有更好的去噪效果.
To remove the trailing noise, histogram fuzzy based filter denoising methods often have the problems of image blurring and residual noisy. To address this problem, the authors of this paper propose a new image de noising method based on Generalized Gaussian Mixture （GGM） model and weighted average image filter. Firstly, the generalized Gaussian mixture model for image is constructed. Secondly, the noise data is determined accord ing to the feature differences between this point and its neighbors. Finally, a weighted average filter is construct ed by the GGM to build an image denoising. Histogram based filter and classical partial differential equation method are compared with the proposed method. Experimental results show that the method has a better denois- ing effect than the other methods.
出处
《常熟理工学院学报》
2012年第8期89-93,共5页
Journal of Changshu Institute of Technology
基金
江苏省高校自然科学基金项目"基于灰色粗糙集理论的不确定性信息处理技术研究"(10KJB520004)
常熟理工学院项目"基于网络坏境的计算机专业自主学习模式研究"(CITJGGN201117)
关键词
广义高斯混合模型
拖尾噪声
加权平均滤波器
Generalized Gaussian Mixture model
trailing noise
weighted average filter
