InSAR interferogram filtering methods in the contourlet domain

Abstract

The interferogram contains much noise which reduce the precision when phase unwrapping. In this paper, we filter the interferogram in the contourlet domain. The contourlet transform (CT) has flexible aspect ratios and can effectively capture geometry information of interferogram edges. However, the CT is lack of the feature of translation invariance. Hereby we study the cycle-spinning CT (CSCT) to convert the commonly CT to translation invariance. Firstly, we translate the original interferogram before being decomposed. Secondly, we decompose the translated interferogram using the CT, and modify the coefficients. Finally, we reconstructed the interferogram with the modified coefficients and translate back. In the experimentation, the data is selected both from the plain and mountain area. The results show that the CSCT outperform the discrete wavelet transform (DWT), the CT in terms of the residues number and the mean value of the correlation coefficient. In texture retrieval, the CSCT shows improvements in performance for various oriented texture and the results indicate a better compromise between noise removal and the detail preservation. Besides, in the mountain area, the CSCT performed well than in the plain area because there is more texture in the mountain area.