An Automatic Threshold Selection Approach for Remote Sensing Imagery of Multimodal Histograms
An Automatic Threshold Selection Approach for Remote Sensing Imagery of Multimodal Histograms
复制标题
DOI：
--
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
Li Yi-ku
中科院分区：
文献类型：
--
作者：
Li Yi-ku
针对多峰直方图的遥感图像，提出了一种改进的自动阈值选取方法。首先，利用均值滤波对多峰直方图进行平滑处理，消除直方图中的小振荡，保留反映直方图整体趋势的大尺度波峰和波谷；其次，通过搜索直方图中的局部极大值(峰)和局部极小值(谷)来确定合理的阈值。该算法只要求用户确定应保留在直方图中的合理峰值个数。对于大比例尺遥感图像，计算的阈值优于大津法和双峰法的计算结果。此外，该算法还具有较高的实用价值通过TM影像水体自动提取的实验表明，该算法能够有效、高效地批量提取水体、冲积扇等遥感信息。
An automatic threshold selection approach is enhanced for remote sensing imagery of multimodal histograms.Firstly,the mean filter is utilized to smooth the multimodal histogram,eliminate the small oscillation in the histogram and keep the large scale peaks and valleys,which reflect the overall trend of the histogram.Secondly,the reasonable threshold is decided by searching the local maxima(peak)and local minima(valley)in the histogram.The algorithm only requires the user to determine the reasonable number of peaks,which should be kept in the histogram.For the large scale remote sensing imagery,the calculated thresholds are better than those calculated by Otsu method and bimodal method.Moreover,the computational complexity of the algorithm is relative low and it is suitable for fast threshold selection.The experiment on automatic extraction of water body in TM imagery shows that the proposed algorithm can effectively and efficiently batch extract remote sensing information such as water body,alluvial fan,etc.