干涉雷达成像之SLC影像是未经任务预处理的，斑点噪声将影响到强度和相位两个方面。但是，通常的滤波方法主要针对幅度影像。照搬这些方法很可能会影响到实际的相位值。合理的滤波方法对于后续的数据处理是非常重要的。该文从干涉成像机理出发提出了一个中值-自适应平滑滤波的解决方案，避免了估计局部地形的复杂计算，又可获得满意的去噪效果。

A new adaptive approach combined with Media filter and gradient-based adaptive filter is presented for removing the noise in the interferogram. Filtering phase noise in an interferogram is an important aspect in INSAR data processing. But any improper altering of the wrapped phase may influence the quality of derived DEM because the interferometric phase contains the topographic information. Therefore one of the difficulties in phase noise filtering is how to remove the noise and preserve the sharp sawtooth profile of the fringe effectively. The interferometric phase image behaves with the fringe in the form of modulus 2πwrapped phase. Filtering directly in interferogram may smooth the sharp edge of the sawtooth in the fringe and damage the unwrapped phase contribution. Some investigators have shown that the processing should not be conducted directly in the phase image，but in the complex plane. With the filtering presented in this paper is done in the complex plane before the phase image is formed by means of the arc tangent operator，i. e. in the real and imaginary component of the interferogram respectively. The gradient-based filtering is an iterative algorithm. In the iteration，the moving window average with adaptive weight is conducted. The weight is varied with the gradient contribution of phase image. The less of the weight is corresponding to the larger gradient value so that the boundary of the fringe is not blurred. Additionally，the moving window is small to avoid the average of the phase values belonging to the different sides of the boundary. But the algorithm is not effective for the spike-like noise because it may also arose the great gradient and be treated as the boundary. To suppress the spike-like noise，the small window Media filtering is applied before the gradient-based average. Compared with existing approach for phase noise filtering，the presented approach does not need complex procedure for the parameter estimation and are easy to be implemented. The effectiveness is verified with the interferogram generated from SIR-C/ X-SAR repeat-pass INSAR data.