摘要
在复杂地表类型的图像变化检测研究中,纹理特征变化能有效地反映不同时期地表的变化情况。基于此,研究纹理特征增强的变化检测方法,采用方向对数变差函数算法对图像的纹理特征进行增强,使用基于对数变差函数的结构相似度算法提取多尺度图像中不同频段的纹理差异特征。通过基于关联度的模糊C均值聚类改进方法对差异特征进行分类,并利用区域生长优化分类结果,提取最终的变化检测结果。实验结果表明,与NSCTKFCM和UDWTKMEAN等方法相比,该方法检测结果精确,提取的变化区域边缘结构清晰,同时减少了光照和噪声的干扰。
In the study about the change detection of complex land surface,the change of texture feature can reflect the characteristics of land surface in different periods. This paper focuses on the methods to detect the change of the texture features,it adopts the improved method of variogram to reinforce the texture features,and uses the Stractural Similarity of Variogram( VSSIM) to extract the texture difference in different frequency bands of multi-scale image. A fuzzy C means clustering method based on relation degree is presented to classify the differences,and the regional growth is used to optimize the classification so as to achieve the final results. Compared with traditional method such as NSCTKFCM and UDWTKM EAN in the references,experimental results show that the detect result of the method is more precise,the extracted regional edges are more clear,and the influence caused by light and noise is reduced simultaneously.
出处
《计算机工程》
CAS
CSCD
北大核心
2016年第2期224-228,共5页
Computer Engineering
基金
国家"973"计划基金资助项目(613XXX0301)
中国博士后科学基金资助项目(2013M542467)
关键词
纹理增强
变化检测
变差函数
结构相似度
关联图
texture enhancement
change detection
variogram function
Structural Similarity(SSIM)
correlation diagram