基于双边滤波与小波阈值法的矿区遥感图像处理

摘要

矿区遥感图像为开采沉陷监测与预计提供了大量的精确数据,对于实现矿区开采沉陷高精度监测与预计发挥了重要作用.由于矿山遥感图像受到矿区复杂的自然环境、成像系统电压不稳定等因素的干扰,导致实地获取的矿山遥感图像在很大程度上受到噪声干扰,图像整体较模糊,难以准确提取有用信息.为此,将双边滤波算法与小波阈值去噪算法相结合,提出了一种矿区遥感图像去噪算法.鉴于双边滤波算法具有较好的边缘保持性能,故首先采用该算法对矿区遥感图像进行处理,以滤除一部分噪声;然后对初次滤波后的图像进行3层小波分解,对分解得到的低频分解系数不作处理,对高频分解系数提出了一种小波域指数型阈值去噪模型进行滤波;最后通过低频分解系数与滤波后高频分解系数重构,获得了高清晰度的矿山遥感图像.采用某矿区2幅遥感图像进行试验,并采用双边滤波算法、小波软阈值去噪模型与新算法进行试验对比,结果表明:新算法在去除噪声的同时,能够最大限度地保持图像细节信息的完整性,其性能总体优于其余2类算法,对于高效处理矿山遥感图像,进而提高矿山开采沉陷监测与预计精度有一定的参考价值.%A large number of data for the mining subsidence monitoring and prediction is provided by the remote sensing image of mining area,it plays an important role of the mining subsidence monitoring and prediction with high precise of mining area. Due to the existing factors of the complex of the natural environment of mining area,instability of image system voltage, the remote image of mining area is usually influenced by the above factors,the overall is blurred,so,it is difficult to extract the useful information accurately. Combing with the bilateral filtering algorithm and wavelet thresholding method, a new filtering method of the remote sensing image of mining area is proposed. Firstly,based on the pefect performance for edge preserving of bilateral filtering algorithm,so,it is used to deal with the noise remote sensing image of mining area to eliminate the part of the noise;then,the initial filtering image is conducted three-layers wavelet decomposition, the low-frequency and high-frequency decomposition coefficients are obtained,because of the low-frequency coefficients do not contain noise,so,keep it unchanged,a new wavelet exponential thresholding model is proposed to deal with the high-frequency wavelet decomposition coefficients;fi-nally,the filtered high-frequency wavelet decomposition coefficients and the original low-frequency wavelet decomposition coef-ficients are reconstructed,the filtered remote image with good visual effects is obtained. The two remote sensing image of a min-ing area is taken as the test data,the test results show that the filtering effects of the new filtering method proposed in this paper is superior to the bilateral filtering algorithm and wavelet soft thresholding model,it has some reference for processing the re-mote sensing of mining area with high efficient and improving the precise of mining subsidence and prediction.