Abstract: The existing factors of uneven illumination,coal dust and circuit voltage instability of video surveilance image acquistition devices,resulting in a lot of noises are existed in video surveilance image,the accurate interpretation of mine all kinds of production information is affected.Combined with discrete wavelet transform (DWT) and improved median filtering algorithm,a filtering algorithm of mine video surveilance with high efficiency is proposed.Firstly,according to the distribution characteristics of the noise in mine video surveilance image,the adaptive noise detection operator is proposed,according to the noise detection results,the improved median filtering algorithm is adopted to filtering out the salt & pepper noise in mine video surveilance image;then,the filtered image is conducted three-layers discrete wavelet transform,the high-frequency and low-frequency coefficients are obtained,the gaussian noise is not distributed in low-frequency coefficients,most of the gaussian noise is distributed in high-frequency coefficients,so,the low-frequency coefficients can be unchanged;finally,the a improved wavelet thresholding filtering function model is proposed to filtering out the gaussion noise in high-frequency coefficients,the filtered high-frequency coefficients and original low-frequency coefficients are reconstructed,the high resoulution image after denoising is obtained.The mine video surveilance images of a mine of Lu′an city,Shanxi province are obtained,the filtering effects of wavelet thresholding filtering function model,median filtering algorithm and the algorithm proposed in paper are analyzed,besides that,the indicators of signal noise ratio (SNR) and algorithm operation time are adopted to evaluation the effects of the above algorithms,the results show that,the performance of the algorithm proposed in this paper is better than others,besidesthat,the operation time of the algorithm proposed in this paper is also shorter than others.

Key words: Mine video surveilance image, Discrete wavelet transform, Median filtering algorithm, Noise detection, Wavelet thresholding filtering function model