基于多Gamma分布模型的SAR图像直方图分割算法  

SAR histogram segmentation algorithm based on multi-gamma model distribution

作　　者：文政颖[1] 于海鹏[1] 

机构地区：[1]河南工程学院计算机学院,河南郑州450007

出　　处：《计算机工程与设计》2014年第6期2104-2108,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61142010)

摘　　要：为了提高SAR图像的分割精度,提出了一种基于多Gamma分布的SAR图像直方图分割算法。采用极大似然技术估算直方图参数,通过二阶微分估算了直方图模型数目,利用模型Gamma分布估算了SAR图像视数;用最大Gamma函数估算了直方图参数;利用分区最小分类误差原理计算了各个分区的阈值。对EnviSat ASAR图像(512×512)进行了分割实验和比较,结果表明该算法使得SAR图像的边缘特征得到了极大保留,且图像中的斑点噪声现象也得到了有效抑制,比传统的阈值分割算法(阈值为40)具有更高的分割精度,可为SAR图像的特征识别和变化检测提供更好的方法。To improve the segmentation accuracy of SAR image,a new algorithm for segmentation of SAR images based on multi-Gamma model distribution used in histogram was proposed.The maximum likelihood technique was used to estimate the histogram parameters,and the second derivative of the histogram was used to estimate the number of the histogram models.The number of looks of SAR image was estimated using Gamma distribution model,and the histogram parameters were estimated by the maximum of the Gamma function,at last,the threshold for each partition is calculated using the zoning minimum classification error principle.Segmentation experiments and comparison results on EnviSat ASAR image （512 × 512） showsed that SAR image edge feature was greatly reserved by the algorithm,and the phenomenon of speckle noise in the image was effectively suppressed.It had higher accuracy of segmentation than the traditional threshold segmentation algorithm,which could provide a better method for feature recognition and change detection of SAR image.

关 键 词：SAR图像 GAMMA分布 改进直方图 极大似然法 阈值分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]