矿井视频监控由于受到井下光照不均匀、高浓度粉尘、成像器件电压不稳定等因素的影响,导致实时获取的图像对比度不高且含有大量的颗粒状噪声,严重影响对图像的准确判读与分析.为此,对经典中值滤波算法进行了改进,提出了一种融合噪声检测策略与投票统计学原理的改进型中值滤波算法.首先对获取的原始矿井视频监控图像采用双平台直方图均衡化算法(Double platform histogram equalization,DPHE)进行处理,该算法通过对图像各灰度级对应的像素点数目设定2个阈值(平台值)对图像背景信息进行抑制,对图像细节信息进行增强;然后将增强后的图像划分为多个规则图像块,根据各图像块内像素点灰度值的分布特征设计了一种噪声图像块自适应检测方法,若某图像块内的灰度值异常点数目超过一定限值,则该图像块被标记为噪声图像块;最后,对噪声图像块进行投票中值滤波.采用C++语言对所提算法进行了编程,并采用2幅采集于某煤矿井下综采工作面的视频监控图像进行试验,结果表明:所提算法对于不同失真程度的矿井视频监控图像的处理效果优于中值滤波、直方图均衡化等算法,对于最大限度地发挥矿井视频监控系统的作用有一定的参考价值.%The mine video monitoring images are usually affected by the factors such as underground uneven illumina-tion,high dust concentration and voltage instability of imaging devices and so on,the quality of the real-time acquisition image is unsatisfactory,to be specific,the image contrast is lowed and a large amount granular noise is existed in the image,so,it is hard to conduct the accurate interpretation and analysis of the image. The classical median filtering algorithm is improved,a new improved median filtering algorithm based on noise detection strategy and voting statistics theory is proposed. Firstly,the double platform histogram equalization algorithm ( DPHE) is adopted to deal with the original mine video monitoring image ob-tained in real-time,based on the pixel numbers corresponding with the gray scale of the image,the two thresholds ( platform values) are set to suppress the image background information and enhance the image detail information;then,the enhanced im-age is divided into a few of image blocks,a noise image block adaptive detection method is proposed based on the distribution characteristics of the gray values of the pixels of all image blocks,if the number of the pixels with gray value anomalies of a im-age block is beyond a certain threshold,it can be marked as noise image block;finally,the noise blocks are filtered by the vot-ing median filtering algorithm. The programme of the algorithm proposed in this paper is written by C++language,the two mine video monitoring images obtained in a working face of a mine are taken as the experimental data,the experimental results show that the performance of the different degree of distortion of mine video monitoring images are superior to median filtering algo-rithm and histogram equalization algorithm,there fore,it has some reference for making full of the mine video monitoring sys-tem.
展开▼
