基于证据理论的小波萎缩图像去噪  

Image denoising by wavelet shrinkage based on evidence theory

作　　者：杨海峰[1] 侯朝桢[1] 

机构地区：[1]北京理工大学信息科学技术学院自动控制系,北京100081

出　　处：《光学技术》2005年第5期713-716,共4页Optical Technique

摘　　要：提出了一种基于D-S证据理论的小波萎缩图像去噪方法。对含噪图像进行离散平稳小波变换后,运用Bayes方法分得各层高频子带的小波萎缩系数,根据小波萎缩系数的空间及层间相关性,利用D-S证据理论的合成法则对初始小波萎缩系数进行融合修正。实验结果表明,该方法在有效地去除图像中的噪声的同时,还能较好地保留图像的边缘信息。算法在性能指标和视觉质量上均优于Donoho的小波软阈值去噪方法、传统的中值滤波法和Winner滤波法。A new method for image de-noising by wavelet shrinkage based on evidence theory was given. In the method, a noise image was multi-scale decomposed by discrete stationary wavelet transform. Bayes method was used to gain original wavelet shrinkage factors of high frequency subbands. According to scale and space consistency of original wavelet shrinkage factors, wavelet shrinkage factors were modified by fusion rules of D-S evidence theory. The experiment result shows that new method can not only effectively get rid of noise but also preserve edges of image well. As to performancy and visual quality, the algorithm is better than the wavelet soft-threshholding given by Donoho, the traditional midian value fliting method and winner fliting method.

关 键 词：D-S证据理论 图像去噪 平稳小波变换 

分 类 号：TP911.7[自动化与计算机技术]