A new method for SAR image coregistration from interferometric coherence modeling

摘要

A new method on single-looking complex SAR images coregistration from interferometric coherence modeling is proposed to accurately seek homologous points. The set of Doppler, slant-range and ellipsoid equations are employed to compute the coarse matching position. The mathematical model of spatial coherence surface can be established using discrete correlation values of central pixel and its adjacent positions. The accurate matching pixel of homologous point can be derived through the maximization operation of modeled mathematical coherence surface. The proposed interferometric coherence modeling method can greatly mitigate the coherence loss due to discrete searching windows in traditional SAR image coregistration algorithms. The ALOS/PALSAR acquisitions over the Longmenshan mountainous area in southwestern China are employed in the study. The testing results show that the coregistration accuracy of SAR images can reach (0.1～.01) pixels. The statistical interferometric coherence is improved with almost 5%. The quality improvement of derived interferogram validates the reliability of the proposed SAR image coregistration method.