基于小波阈值法的矿山遥感图像非局部均值去噪

Non-local Means Filtering Algorithm of Mine Remote Sensing Image Based on Wavelet Thresholding Method

矿山遥感图像在获取、压缩、传输、解码的过程中易混入大量随机噪声,导致图像清晰度较低,难以直接进行分析研究。为此,将小波变换与非局部均值滤波算法(Non-local means filtering algorithm,NLM)相结合提出了一种基于小波阈值法的矿山遥感图像非局部均值去噪算法。该算法首先结合小波硬阈值、软阈值去噪模型以及现有的改进型小波阈值去噪模型的特点,建立一种改进型小波阈值去噪模型并用于去除遥感图像中的随机噪声;然后将原始遥感图像与小波去噪后的图像作差运算,得到原始差值图像,再对原始差值图像进行非局部均值滤波,得到滤波后的差值图像;最后将小波去噪后的图像与滤波后的差值图像进行融合。采用MATLAB语言编写程序,试验数据为白云鄂博矿区的2幅遥感图像,采用峰值信噪比(Peak signal noise to ratio,PSNR)和算法耗时等指标对算法去噪效果进行评价。试验结果表明:所提算法的去噪结果明显优于小波软阈值去噪模型及非局部均值滤波,此外,该算法耗时相对于其余2类算法而言也有一定的优势,对于提高矿山遥感图像的判读精度有一定的参考价值。

A large number of random noise is added into the mine remote sensing image in the process of acquisition,compression,transmission,decoding of mine remote sensing image,which resuling in the decressing of the quality of mine remote sensing image,so,it is difficulty to analyze the mine remote sensing image directily. Combing with the wavelet transform and non-local means filtering algorithm,a new mine remote sensing image non-local means filtering algorithm based on wavelet thresholding method is proposed. Firstly,a new improved wavelet thresholding denoising model is proposed based on the characteristics of hard wavelet thresholding denoising model,soft wavelet thresholding denoising modela and the existed improved wavelet thresholding denoising models,and it is used to deal with the random noise exieted in the mine remote sensing image;secondly,the difference image of the orginal mine remote sensing image and the image filtered by the improved wavelet thresholding denoising model is obtained;thirdly,the difference image is filtered by the non-local means filtering algorithm,the filtered difference image is obtained;finally,the image filtered by the improved wavelet thresholding denoising image and the filtered difference image are integrated. The programme of the filtering algorithm proposed in this papaer is obtained by adopting MATLAB language,the two remote images acquired from Bayan Obo mining area are used as the experiment data,the index of peak signal noise to ratio( PSNR) and time-consuming is adopted to evaluate the filtering effects of the filtering algorithm mentioned in this paper. The experiment results show that the performance of the algorithm proposed in this paper is better than the soft wavelet thresholding denoising model and non-local means filtering algorithm,besides that the time-consuming of the algorithm proposed in this paper is shorter than the other two algorithms. Therefore,the algorithm proposed in this paper has important reference value for improving the interpretation precision of mine remote sensing image.