矿山遥感图像小波域模糊隶属度阈值去噪算法

矿山遥感图像作为一类重要的矿山测量数据,其质量在很大程度上受到矿区成像环境的影响。鉴于小波变换的图像多尺度分析特性,提出了一种小波域遥感图像模糊隶属度阈值去噪算法。该算法首先对失真的矿山遥感图像采用均值滤波算法进行预处理,消除一部分图像噪声。然后对预处理后的图像进行3层小波分解,对于基本不受噪声污染的低频小波分解系数不作处理,根据高频小波分解系数的噪声分布特征,设计了模糊隶属度阈值去噪模型用于去除其中的噪声,该模型对小波软阈值去噪模型进行了2点改进:1根据图像局部区域噪声信息与非噪声信息难以有效区分的情况,设计了模糊隶属度因子,通过设定特定的小波阈值,对不同的高频小波分解系数是否含有噪声进行自适应判定;2顾及到图像小波分解层数以及各高频小波分解系数的幅值,对经典小波贝叶斯阈值计算方法进行了改进。最后将原始低频小波分解系数与去噪后的高频小波分解系数进行重构,得到高清晰度的矿山遥感图像。采用山东兖州矿区的1幅遥感图像进行试验,并引入了边缘保持指数(Edge protection index,EPI)、信噪比(Signal noise to ratio,SNR)进行去噪效果评价,结果表明:所提算法对于失真的矿山遥感图像的去噪效果明显优于小波硬阈值、软阈值去噪模型,并且相对于已有的2类改进型去噪模型,优势也较显著。

Fuzzy Membership Degree Threshold Denoising Algorithm of Mine Remote Sensing Image in Wavelet Domain

Mine remote image is an important kind of mine surveying data,its quality is affected by the imaging environ-ment in mining area to a large extent. According to the image multi-scale analysis characteristics of wavelet transform,a new fuzzy membership degree threshold denoising algorithm in wavelet domain is proposed. Firstly,the mine distortion remote sens-ing image is processed by average filtering algorithm to filter out a part of the noise distributed in the mine distortion remote sensing image. Secondly,the pre-processing image is conducted three-layers wavelet transform,the low frequency wavelet de-composition coefficients that are not polluted by noise are keep unchanged,according to the noise distribution characteristics of high frequency wavelet decomposition coefficients, a new fuzzy membership degree threshold denoising model is designed to deal with the noise distributed in the high frequency wavelet decomposition coefficients,the classical soft denoising model is improved from the two aspects:①it is very difficult to distinguish between the noise information and image information,so,a fuzzy membership degree factor is designed,the high frequency wavelet decomposition coefficients that are polluted by noise that can be determiend effectively by setting specific wavelet threshold;②according to the layer number of wavelet decomposi-tion and amplitude of the high frequency wavelet decomposition coefficients,the classical wavelet Bayesian threshold calculation method is proposed. Finally,the original low frequency wavelet decomposition coefficients and high frequency wavelet decompo-sition coefficients after denoising are conducted reconsitution, the mine remote sensing image with high definition is ob-tained. The remote sensing image is obtained in a mining area of Yanzhou city,Shandong province is taken as the experimental data,edge protection index ( EPI) and signal noise to ratio ( SNR) are used to evaluate the denoising effects of the algorithms in experiment,the experimental results show that the the denoising ability of the algorithm proposed in this paper is superior to the classical hard and soft wavelet denoising model and its two improved denoising models.