Abstract: Due to the high concentration dust and uneven illumination,it is easy to make the mine video image contains a lot of noises,the mine video images obtained in real-time are vague as a whole,it affect the interpretation and analysis of them,in order to make use of the mine video images as much as possible,it is necessary to conduct mine video images processing.Based on wavelet transform that a effective images analysis method,a new mine video images filtering algorithm in wavelet domain is proposed.Firstly,the mine video image is conducted two-layers wavelet transform,the low-frequency decomposition coefficients and high-frequency decomposition coefficients are obtained,the low-frequency decomposition coefficients and high-frequency decomposition coefficients are conducted inverse wavelet transform respectively,the spatial domain low-frequency image and high-frequency image of the original mine video image are obtained;secondly,according to the characteristics of low contrast and without noises interference basically of spatial domain low-frequency image,the homomorphic filtering algorithm is adopted to conduct image enhancement;then,based on analyzing the characteristics of the non-local means filtering(NLM)algorithm,the similarity weighted calculation method and searching scope of the image blocks of non-local means filtering(NLM) algorithm are improved,a improved non-local means filtering(INLM) algorithm is put forward,it is used to filter out the noises in spatial domain high-frequency image;finally,the enhanced spatial domain high-frequency image and filtered spatial high-frequency image are superimposed.The mine video image obtained in Wangjiazhai coal mine of Xinren county,Guizhou province is taken as the experimental object,the structural similarity(SSIM)and root mean square error(RMSE)are used to evaluate the performance of the above algorithms.The results show that the performance of the algorithm proposed in this paper is superior than the non-local means filtering (NLM) algorithm and its two improved algorithms,it has some reference for the mine video image processing.

Key words: Mine video images, Wavelet transform, Homomorphic filtering, Non-local means filtering, Similarity weighted, Searching scope of image blocks