Distributed coherent aperture radar (DCAR) can be applied to improve the target detection performance through adjusting transmitting/receiving time and phase of spatially distributed radars via the coherent parameters (CPs). Therefore, obtaining high-accuracy CPs is the prerequisite for DCAR to realize coherent synthesis. However, the ideal orthogonal waveforms on common spectral used for estimating CPs do not exist in practice. Cross-correlation energy leakage between the orthogonal waveforms seriously impairs the accurate acquisition of CPs. To solve the above problem, a CPs estimation algorithm is proposed based on reconstructing the “clean” matched filtering outputs via a reconstruction-elimination-econstruction framework, and CPs are estimated via reconstructed “clean” signals. Simulations results show that the proposed algorithm can effectively suppress the cross-correlation energy leakage, and the accuracy of CPs estimates are significantly improved.