Abstract: The paper roundly analyzes the biases of space multi-look InSAR coherence estimation and constitutes the Multi-Look InSAR Error model of the Coherence estimation estimated by Region Growing Theory. Based on this InSAR Error model, the paper proposes a highly accurate iterative estimating method Firstly, it chooses the moving window by region growing theory which ensures pixels in the window belong to the same distribution, then the bias resulting from the pixels belonging to different distributions is decreased .Secondly the bias resulting from too low coherence and too small number of pixels is decreased by using GaussNewton Iteration algorithm. The Iteration algorithm bases on the equation between the multi-look coherence and the real coherence and the equation is educed from the complex Hermit product Speckle Noise Modeling. Experiment results validate the validity of this algorithm