一种基于相干系数的复数图像滤波方法-《计算机技术与发展》

文章信息/Info

Author(s):: ZHAO Shan¹; YANG Tao-li²; 1.School of Remote Sensing Information and Engineering,Wuhan University,Wuhan 430000,China; 2.University of Electronic Science and Technology of China,Chengdu 611731,China

Keywords:: coherence coefficient; complex number; interferometric synthetic aperture radar(InSAR); region growth; self-adaptation

摘要:: 复数数据类型的图像具有较强的条纹性,其中最具代表性的是干涉合成孔径雷达(interferometric synthetic aperture radar,InSAR)图像。以复数图像处理为研究课题,针对现有滤波算法自适应性差,不适应复数数据处理,条纹连贯性低的缺点,提出了一种新的基于相干系数的滤波方法。从相干系数的选择和区域生长方法两个方面进行了改进。相干系数是复数图像对相似度的重要度量指标,该算法以相干系数作为滤波窗口选择的指标,并通过设定一定的阈值作为生长条件,形成滤波模板。由此得到的滤波窗口的形状、大小更为灵活,和实际条纹的相似度达到89%以上。经过该方法处理后的图像的边缘保持指数和等效视数均优于圆周期均值滤波、中值滤波,最小均方差滤等指标。可知相干系数滤波算法的滤波效果最好,图像条纹的连贯性得到有效的补偿。文中实现了复数图像中同质像素的选择,复数数据的处理以及相位保真的优化。

Abstract:: The complex number images have strong fringe,among which the most representative one is interferometric synthetic aperture radar(InSAR) image. Taking complex imageprocessing as theresearch subject,wepresent anewfiltering method based on coherence coefficients to overcome the disadvantages of existing filtering algorithms,such as poor self-adaptability,unadaptability to complex number processing and low stripe coherence. This method is improved from coherence coefficient selection and region growth method. Coherence coefficient is an important measure index of similarity of complex image. This algorithm takes coherence coefficient as the index of filter window selection and forms filter template by setting a certain threshold as the growth condition. The result shows that the shape and size of the filter window are more flexible,and the similarity with the actual fringes is up to 89%. The edge preservation index and equivalent view number of the image processed by this method are better than those of the circular mean filter,the median filter,and the minimum mean square filter,etc. It can be seen that the coherence coefficient filtering algorithm is satisfactory and the coherence of image fringes is effectively preserved. The selection of homogeneous pixels,the processing of complex number and the optimization of phase fidelity in complex number images are realized.