中文摘要：

矿井视频监控系统的相当一部分图像信息采集传感器处于低照明、高密度粉尘的环境中,导致获取的图像出现忽明忽暗、噪声较多的现象,很大程度上干扰了对井下生产状况的有效监控。为此,将小波阈值去噪算法与非局部均值滤波算法（Non-local means filtering,NLM）相结合,提出了一种井下视频图像去噪算法。该算法对获取的原始矿井视频图像进行单层小波变换,对得到的低频系数和高频系数分别进行如下处理：1将低频系数进行单层小波变换,得到次低频系数1和次高频系数1,对次高频系数1采用改进型小波阈值去噪模型进行噪声抑制后与次低频系数1进行重构,得到低频图像;2将高频系数进行单层小波变换,得到次低频系数2和次高频系数2,对次高频系数2予以舍弃,对次低频系数2采用小波软阈值去噪模型处理后进行系数重构,得到高频图像。对获取的低频、高频图像进行融合,并对融合后的图像进行非局部均值滤波,得到高清晰度的矿井视频图像。采用VB语言对所提算法进行编程试验,并与小波硬阈值去噪模型、小波软阈值去噪模型、非局部均值滤波算法进行试验对比,结果表明：该算法去噪后的矿井视频图像清晰度以及峰值信噪比（Peak singnal noise to ratio,PSNR）、均方根误差（Root mean square error,MSE）等指标明显优于其余3类算法。

英文摘要：

A large number of the image information acquistition sensors of the mine video monitoring system are placed in the environment with the characteristics of low light,high density of dust,which is resulted in the images obtained by the image information acquistitions with the characteristics of low contrast and high density of noise,so,the effective monitoring of mine production is influenced to some extent. Combing with the wavelet thresholding algorithm and non-local means filtering algorithm（ NLM）,the filtering algorithm of mine monitoring video image is proposed. The obatined original mine monitoring video image is conducted single-layer wavelet transform,the obatined low-frequency and high-frequency coefficients are processed as following： 1the low-frequency coefficient is conducted single-layer wavelet transform,the secondly low-freqency coefficient 1and secondly high-frequency coefficients 1 are obtained,the secondly high-frequency coefficients 1 area filtered by the improved wavelet thresholding denoising model proposed in this papaer,the filtered secondly high-frequency coefficients 1 are reconstructed,the low-frequency image is acquired; 2the high-coefficients are conducted single-layer wavelet transform,the secondly low-frequency coefficient 2 and secondly high-frequency coefficents 2 are obtained,the secondly high-frequency coefficients 2 are set to zero,the secondly low-frequency coefficients 2 are processed by the soft wavelet thresholding denoising model,the filtered secondly low-frequency coefficients are reconstructed to obtain the high-frequency image. The above low-frequency and high-frequency image are intergrated,the processed mine monitoring video image wih high resolution. The programme of the algorithm proposed in the papaer is writed by VB language,the experimental results show that the resolution of the image processed by the algorithm proposed in this paper is higher to the ones of the hard wavelet thresholding denoising model,soft wavelet thresholding denoising model and non-local