Improved SNR Optimum Method in POLDINSAR Coherence Optimization

Abstract

The traditional methods for coherence optimization in the framework of multibaseline (MB) polarimetric differential interferometric synthetic aperture radar (DInSAR) applications, such as Best, MB1 equal scattering mechanism (MB1-ESM), suboptimum scattering mechanism (SOM), and exhaustive search polarimetric optimization (ESPO), all have some disadvantages. The MB2-ESM method just can be preferred for its normal accuracy and fast computing time. Signal–noise ratio optimum (SNR-OPT) in the coarse grid followed by the conjugated gradient method (SNR-OPT-CG-CGM) can be selected because of its higher accuracy and acceptable cost time. SNR-OPT has higher computational efficiency compared with ESPO because it makes the 4-D coherence optimization problem transform into two independent 2-D optimization problems (“2 $+$ 2” optimization problem). However, SNR-OPT still costs much time. In this letter, we propose a new method which can further make this “2 $+$ 2” optimization problem transform into one 2-D and two independent 1-D optimization problems (“ $2+1+1$ ” optimization problem). Thus, the computational efficiency will be improved much more compared with SNR-OPT, and meanwhile, the accuracy just decreases a little. Seven full polarimetric RADARSAT-2 images are taken for experiment, and the results also show that the improved SNR-OPT-CG-CGM method is a better method considering the tradeoff between computation time and accuracy compared with other methods for DINSAR applications.