Satellite-borne SAR (synthetic aperture radar) is for high-resolution geosurface measurements. Recently, a feature extraction method based on CCD (coherent change detection) was developed, where a slight surface change on the geosurface is detected using the phase relationship between sequential complex SAR images of the same region made at different times. For accurate detection of the surface change, the log-likelihood method has been proposed. This method determines an appropriate threshold for change detection, making use of the phase characteristic of the changed area, and thus enhances the detection probability. However, this and other conventional methods do not seek to proactively employ phase information of the estimated coherence function, and their detection probability is often low, especially in the case that the target has small surface or local uniform changes. To overcome this problem, this paper proposes a novel transformation index that considers the phase difference of the coherence function. Furthermore, we introduce a pre-processing calibration method to compensate the bias error for the coherence phase which resulting mainly from the orbit error of the antenna platform. Finally, the results from numerical simulations and experiment modeling of the geosurface measurement verify the effectiveness of the proposed method, even in situations with low SNR (signal to noise ratio).