基于多重滤波的高分辨率影像中线性特征提取

在高分辨率遥感影像中，线性特征提取一直是个难题，难主要表现在线性地物的边缘信息和噪音是一对矛盾体，边缘信息越明显，噪音就越强。本文利用高通滤波和统计滤波器，构建一个多重滤波，对IKONOS影像进行处理，通过三种不同高通滤波窗口对IKONOS影像处理结果的比对，得到一个较好的高通滤波处理窗口，处理后的IKONOS影像边界特征突出、区域模糊程度减小，目标地物大致轮廓已经区分出来，在此基础上对影像进行统计滤波处理，噪音明显减少，灰度分布更加集中和平滑，对高通滤波和统计滤波处理后的图像进行边缘检测，经过统计滤波后的图像较只进行高通滤波后的图像检测结果较好。实验结果表明：滤波后的影像噪音明显降低，信噪比明显提高，统计滤波后的影像，能够提高检测精度，表明了从高分辨率卫星影像上自动提取矢量信息的可行性，并为边缘检测的细化与跟踪处理提供了可靠的数据来源。

Linear Features Extraction in the High-resolution Images Based on Multi-filter

It is quite difficult, in high-resolution images, to detach the boundary information of linear surfaces. Used the high-pass filter and statistical filter to build a multi-filter, our research draws the boundary feature of linear surfaces in IKONOS image. The experimental results indicate that boundary information of linear surfaces is clearer processed by statistical filtering than only by high-pass filtering. Meanwhile, noises in thefiltered image are significantly decreasing and the SNR is markedly increasing. Therefore, it offers more dependable data sources for tracking linear surfaces boundary information.