矿井视频图像小波域改进非局部均值滤波  

Improved Non-local Means Filtering Algorithm in Wavelet Domain of Mine Video Images

作　　者：王平均[1] 王伟[1] 

机构地区：[1]海南软件职业技术学院电子工程系,海南琼海571400

出　　处：《金属矿山》2016年第3期143-146,共4页Metal Mine

摘　　要：由于井下粉尘较多,光照不均匀,易导致矿井视频图像中含有大量噪声,实时获取的矿井视频图像整体较模糊,影响了对其进行分析判读。对此,基于小波变换,提出了一种小波域矿井视频图像滤波算法。该算法首先对矿井视频图像进行2层小波变换,对获得的低频和高频小波分解系数分别进行逆小波变换,得到空间域原始图像的低频图像和高频图像;其次根据低频图像对比度较低、基本不受到噪声干扰的特点,对其采用同态滤波算法进行增强处理;然后,在对非局部均值滤波(Non-local means filtering,NLM)算法特点分析的基础上,分别从相似性权重计算、图像块搜索范围自适应确定等方面对其进行了改进,提出了一种改进型非局部均值滤波算法(Improved non-local means filtering,INLM),采用该算法对高频图像进行去噪处理;最后将增强后的低频图像和去噪后的高频图像进行叠加,得到质量较高的矿井视频图像。采用一幅贵州省兴仁县王家寨煤矿井下视频图像进行试验,并将文中所提算法与非局部均值滤波算法及其2类改进型算法进行性能对比,采用结构相似度指数(Structure similarity,SSIM)、均方根误差(Root mean square error,RMSE)等指标对各算法性能进行评价,结果表明:新算法对于矿井视频图像的处理效果优于其余算法,对于高效处理矿井视频图像有一定的参考价值。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 highfrequency 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

关 键 词：矿井视频图像 小波变换 同态滤波 非局部均值滤波 相似性权重 图像块搜索范围 

分 类 号：TD672[矿业工程—矿山机电]