Use of multiple filters for Noise Reduction in InSAR Images
This paper proposed a new filtering technique to reduce noise in Synthetic Aperture Radar (SAR) interfermetric phase images. A general two-stage filter structure is described into which any number of suitable filters can be incorporated. This technique involves the use of multiple filters to eliminate noise from the original signal and reintroduce some of the details that are lost in this process back into the signal.
In this experiment, a pivoting median filter with an adaptive window size followed by a Lee filter with 16 directional windows are used as Stage I filters to reduce noise with minimum loss of detail. In Stage II, a pivoting median filter with a fixed window size (15x15) followed by another Lee filter is chosen for a more aggressive noise reduction. In this stage, loss of detail is tolerated to a larger degree in order to ensure noise is not reintroduced into the output. The filters used in the structure above can be changed, removed or new filters can be introduced to improve the performance as required.
The result is obtained by using the ERS-1/2 tandem mission over Appin, Australia. The performance comparison is done based on the number of residues present in the de-noised output and the RMS errors between the ground truth and the obtained DEM. The results show that the new technique outperformed other widely used interferogram filtering techniques.