InSAR Image Denoising Filter for Accurate DEM Generation

InSAR image noise has a great effect on the efficiency of the phase unwrapping process and consequently the correct generation of the digital elevation model (DEM). However, phase unwrapping can be easier and more efficient by excellent interferometric phase image filtering which can be achieved by high noise reduction, perfect preservation of the image detail, and decreasing the number of residues. In this paper, a grey-scale soft morphological filter is optimized using the genetic algorithm and used to filter out the interferometric phase image noise. The filter parameters are optimized to achieve an optimum balance between the amount of noise reduction and the degree of preservation of the image detail. Simulated and real interferograms are employed to evaluate the performance of the optimized filter. The evaluation is based on both objective and subjective measures. The results demonstrate that the proposed filter achieves high noise reduction with perfect image detail preservation and very small number of residues. This outstanding performance guarantees efficient phase unwrapping and hence accurate DEM generation. The results show also that the filter outperforms other traditional filters used for denoising interferogram images.