中文摘要：

根据窗口内周围各像素与中心像素之间的异向梯度及像素相干值构成中心像素迭代公式中的权函数,提出一种保持条纹细节的干涉条纹图滤波方法。以模拟数据和意大利Etna火山、土耳其地震的真实干涉图数据为例,将新方法与空间自适应滤波、Lee滤波和最优化方向融合方法进行比较。研究结果表明：新算法能够滤除大量噪声同时保持条纹边缘,模拟干涉图和2幅真实干涉图残余点去除率分别达到99.3%,90.3%和87.9%,更有利于后续的相位解缠。

英文摘要：

The kernel pixel’s iteration formula was weighted by coherence and anisotropic gradients between surrounding pixels of the window and the kernel pixel. By this means a new filter for interferogram was presented, which could keep the fringe detail as well. Taking the simulated data and actual interferograms of Etna volcano in Italy and earthquake in Turkey for example, this new algorithm was compared with complex space domain adaptive filter, Lee filter and optimal integration-based adaptive direction filter. The results show that this new method performs well not only in noise reduction but also in preservation of the sharp sawtooth profile of fringe. The residual points reductions in simulate and actual interferograms are 99.3%,90.3% and 87.9% respectively, in favor of following phase unwrapping.