![](https://csdnimg.cn/release/download_crawler_static/85095489/bg1.jpg)
Series No.478
April 2016
金 畿 母
METAL MINE
总 第 478期
2016年第 4期
矿 山遥感 图像 自适 应加权 改进 中值滤波算 法
姚 薇 。 钱玲玲
(1.淮安信 息职业技 术 学院 电气工程 系,江 苏 淮安 223003;
2.江苏电子产品装备制造工程技术研 究开发 中心 ,江苏 淮安 223003)
摘 要 矿山遥感 图像 的质量受成像 区域环境 的影 响较大 ,加之 成像 器件 固有 的缺 陷 ,易 导致获 取的 图像 存在
一
定程度 的失 真 ,影响对遥感 图像的判读和分析。为此,在中值滤波 (Median filtering,MF)算法的基础上进行 了加权
改进 ,提出 了一种适合 于遥感 图像 处理 的 自适应 加权 改进 中值 滤波算 法 (Adaptive weighted improved median filtering,
AWIMF)。首先将遥感 图像划 分为多个尺寸为 7×7的图像块 ,对 每个 图像块 分别统计像 素点灰 度极大 (小 )值 ,将 该
类像素点标记 为第 1类 疑似噪声点 ;其次计算每个图像块的像素点灰度 中值 ,将各 图像块 中的每个像素点灰度值 与对
应的像素点灰度 中值作差 ,将较 大差 值对应的像素点标记 为第 2类疑似 噪声点 ,将 2次检测均被标记为疑似噪声点的
像素点确定 为噪声点 ;然后 ,以每个噪声点 为中心选 取尺寸为 5x 5的滤波窗 口,根据滤波窗 口内各像素点灰度值与所
在滤波窗 口的像素 点灰 度中值的差值计算权重 ,分 别将各 像素 点灰度 值与对 应 的权 重进行 加权计算 ,实 现对噪 声点
的高效滤波 ;最后 ,采用直方图规定 化算法 (Histogram specification,HS)对滤 波后的图像对 比度进 行调 节。采用内蒙
古白云鄂博矿 区遥感 图像进 行试验 ,结果 表明,所提算法 的性 能相对于中值滤波及 其 已有 的几类 改进 型算 法而言,优
势较 明显 。
关键词 矿 山遥感 图像 中值 滤波 疑 似噪声点 加权 改进 中值 滤波 直方图规定化
中图分类号 TD672 文献标志码 A 文章编号 1001.1250(2016).04—101—05
Adaptive W eighted Im proved M edian Filtering Algorithm of M ine Rem ote Sensing Im age
Yao Wei ' Qian Lingling ,
(1.Department ofElectrical Engineering,Huaian College ofInformation and Technology,Huaian 223003,China;2.The Engineering
Technology Research and Development Center ofElectronic Products Equ/pment Manufacturing ofJiangsa Province,Huaian 223OO3,Ch/na)
Abstract The quality of mine remote sensing image is affected by the imaging regional environment greatly,and the ima—
ging device inherent defects make access to the obtained remote sensing image with certain degree of distortion,which appeared
all kinds of random noise.The median filtering(MF)algorithm is weighted improved,a new adaptive weighted improved medi—
an filtering(AWIMF)algorithm that is suitable for remote sensing image processing is proposed.Firstly,the image is divided
into m any image blocks with the size of 7 x7,the maximum and minimum pixel gray values of each image block are counted
out,and the pixels are marked as the class one suspected noise points;secondly,the pixel gray median values of each image
block are calculated,each pixel gray value of the each image blocks and its corresponding pixel gray median values are made
differential operation,the pixels corresponding with larger differential values are marked as the class two suspected noise
points,the pixels that are marked as the suspected noise points in the process of noise detection are identified as noise points;
thirdly,centered with each noise points,the filter windows with the size of 5 ̄5 are selected,based on the each pixel gray values
in the filtering windows and the gray median values of the corresponding filtering windows,the weighted values are calculated,
each pixel gray value and its corresponding weighted value are conducted weighted calculation to filtering out the noise points
effectively;finally。the filtered remote sensing image contrast is adjusted by adopting the histogram specification(HS)algo—
rithm ,the remote sensing image of Baiyun Obo mining area,Inner Mongolia is used as the experimental data,the experimental
results show that the performance of the algorithm proposed in this paper is superior to median filtering(MF)algorithm and its
several existing improved algorithms.
Keywords Mine remote sensing image,Median filtering,Suspected noise point,W eighted improved median filtering,
Histogram specification
收稿 日期 2016-01—07
作 者简介 姚 薇 (1983一 ),女 ,讲师 ,硕士 。
·
101·