We presented a novel approach for accurate temporal coherence decorrelation analysis of InSAR echo signal. The proposed algorithm is divided into three steps: Firstly, using the modified Intensity-Driven Adaptive Neighborhood (IDAN) algorithm to estimate interferometric coherence; then using least squares fitting to remove deviation; finally, obtaining accurate temporal coherence decorrelation component with coherence decomposition technique by separating approximate unbiased coherence. We analyzed the ENVISAT ASAR data from the area of Los Angeles with the new method. The result was compared with the existing methods. The result shows that the new fusion algorithm is able to obtain more reliable and accurate temporal coherence decorrelation component. Meanwhile, the characteristics for the new method of non-threshold and almost adaptive performance were proved.