摘要
针对声呐图像噪声污染严重的问题,在基于形态小波的声呐图像去噪方法中引入了谱聚类算法以实现低信噪比下图像的去噪。给出基于形态中点小波的声呐图像去噪法,在此基础上引入谱聚类的概念,针对谱聚类能快速实现数据分类的特点,对形态中点小波分解后的高频小波系数进行分类,使得包含噪声与细节信号部分的小波系数得以分离。对分离后的两类小波系数分别进行相应的处理,通过与低频小波系数的重构即可得到去噪后的图像。仿真实验表明:该方法在信噪比等性能指标方面均优于其他小波去噪方法,验证了所提方法的可行性和有效性。
Aiming at the serious noise pollution problems in sonar image, spectral clustering algorithm is introduced into denoising method based on morphological wavelet in order to realize image denoising with low SNR. Sonar image denoising method based on morphological midpoint wavelet is given. The concept of spectral clustering is introduced and used for the classification of high-frequency wavelet coefficients after morphological midpoint wavelet decomposition because of its fast data claasification features. The wavelet coefficients including noise and detail signal can be separated. After being seperated, the two kinds of wavelet coefficients are respectively and corresponding processed. The denoised image can be obtained by the reconstruction of lowfrequency wavelet coefficients. The simulation experiments show that the proposed method are better than other wavelet denoising methods and the feasibility and effectiveness of the proposed method are verified.
出处
《传感器与微系统》
CSCD
北大核心
2011年第10期41-43,共3页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(5090902560772128)
关键词
声呐图像去噪
形态中点小波
谱聚类
sonar image denoising
morphological midpoint wavelet
spectral clustering
