矿井视频监控图像改进非局部均值滤波算法

Improved Non-local Means Filtering Algorithm of Mine Video Monitoring Image

矿井成像条件较为复杂,导致视频监控系统所获取的图像往往具有对比度低、且参杂大量随机噪声的特点,给实时监控矿井生产状况带来了不便。为此,采用灰色关联度法改进非局部均值滤波算法,提出了一种矿井视频监控图像改进非局部均值滤波算法。该算法首先对原始矿井视频监控图像采用均值滤波算法进行预处理,得到预滤波图像,分别对原始矿井视频监控图像和预滤波图像划分为5×5大小的图像块,将该2幅图像中对应位置图像块的像素点灰度值集合分别记为待比较序列和参考序列,计算其灰色关联度值,将较小的灰色关联度值对应的原始矿井视频监控图像中的图像块标记为疑似噪声图像块;其次对每个疑似噪声图像块分别统计其像素灰度极大、极小值,并将该类像素点标记为噪声点;然后以每个噪声点为中心取大小为3×3的图像块,进行非局部均值滤波;最后对滤波后的矿井视频监控图像采用直方图均衡化方法进行对比度拉伸,改善图像的视觉效果。试验结果表明:本研究算法无需对图像中每个像素点灰度值进行逐一滤波,提高了图像处理效率,有助于实现矿井视频监控图像的高效处理。

The mine imaging condition is relatively complex,which results that the images obtained by the video monitoring system has the characteristics of low contrast and mixed with large number of random noise,which brought inconvenience to monitoring the mine production conditions in real-time. So,the grey correlation method is adopted to improve the non-local means filtering algorithm,and an improved non-local means filtering algorithm of mine video monitoring image is proposed.Firstly,the original mine video monitoring image is processed by the average filtering algorithm to obtain the filtering image,and the original mine video monitoring image and filtering image are divided into images blocks with the size of 5 × 5,the grey value collections of the pixies in the image bocks with the corresponding position in the above two images are regarded as the compare sequence and reference sequence respectively to calculate the grey correlation value of the compare sequence and reference sequence,and the image blocks with smaller grey correlation values in the original mine video monitoring image can be marked with suspected noise image blocks; secondly,the maximum and minimum grey values of the pixels in the suspected noise image blocks are marked as noise pixels; then,the image blocks with the size of 3 × 3 centered with the noise pixels are processed with the non-local means filtering algorithm; finally,the contrast of mine video monitoring image after filtering can be stretched by using the histogram equalization method to improve the image visual effects. The experimental results show that the algorithm proposed in this paper dose not need to filter the noise pixels one by one,therefore,the filtering efficiency of image processing is improved. It contributes to realize the goal of processing the mine video monitoring image with high efficiency.